Update blis to 1.0

.appveyor.yml

@@ -1,3 +1,5 @@
+skip_branch_with_pr: true
+
 environment:
   matrix:
     - LIB_TYPE: shared
@@ -21,6 +23,12 @@ environment:
       CC: clang
       THREADING: openmp
 
+    - LIB_TYPE: static
+      CONFIG: auto
+      CC: clang
+      THREADING: openmp
+      SANDBOX: yes
+
 install:
 - set "PATH=C:\msys64\mingw64\bin;C:\msys64\bin;%PATH%"
 - if [%CC%]==[clang] set "PATH=C:\Program Files\LLVM\bin;%PATH%"
@@ -32,17 +40,19 @@ build_script:
 - if [%LIB_TYPE%]==[shared] set "CONFIGURE_OPTS=%CONFIGURE_OPTS% --enable-shared --disable-static"
 - if [%LIB_TYPE%]==[static] set "CONFIGURE_OPTS=%CONFIGURE_OPTS% --disable-shared --enable-static"
 - if not [%CBLAS%]==[no] set "CONFIGURE_OPTS=%CONFIGURE_OPTS% --enable-cblas"
+- if [%SANDBOX%]==[yes] set "CONFIGURE_OPTS=%CONFIGURE_OPTS% -s gemmlike"
 - set RANLIB=echo
 - set LIBPTHREAD=
 - set "PATH=%PATH%;C:\blis\lib"
 - set "CFLAGS=-Wno-macro-redefined"
 - bash -lc "cd /c/projects/blis && ./configure %CONFIGURE_OPTS% --enable-threading=%THREADING% --enable-arg-max-hack --prefix=/c/blis %CONFIG%"
 - bash -lc "cd /c/projects/blis && mingw32-make -j4 V=1"
 - bash -lc "cd /c/projects/blis && mingw32-make install"
-- ps: Compress-Archive -Path C:\blis -DestinationPath C:\blis.zip
+- 7z a C:\blis.zip C:\blis
 - ps: Push-AppveyorArtifact C:\blis.zip
 
 test_script:
+# "make checkblas" does not work with shared linking Windows due to inability to override xerbla_
 - if [%LIB_TYPE%]==[shared] set "TEST_TARGET=checkblis-fast"
 - if [%LIB_TYPE%]==[static] set "TEST_TARGET=check"
 - bash -lc "cd /c/projects/blis && mingw32-make %TEST_TARGET% -j4 V=1"

.dir-locals.el

@@ -0,0 +1,32 @@
+;; Emacs formatting for the BLIS layout requirements.
+
+(
+ ;; Recognize *.mk files as Makefile fragments
+ (auto-mode-alist . (("\\.mk\\'" . makefile-mode)) )
+
+ ;; Makefiles require tabs and are almost always width 8
+ (makefile-mode . (
+                   (indent-tabs-mode . t)
+                   (tab-width . 8)
+                   )
+                )
+
+ ;; C code formatting roughly according to docs/CodingConventions.md
+ (c-mode . (
+            (c-file-style . "bsd")
+            (c-basic-offset . 4)
+            (comment-start . "// ")
+            (comment-end . "")
+            (parens-require-spaces . nil)
+            )
+         )
+
+ ;; Default formatting for all source files not overriden above
+ (prog-mode . (
+               (indent-tabs-mode . nil)
+               (tab-width . 4)
+               (require-final-newline . t)
+               (eval add-hook `before-save-hook `delete-trailing-whitespace)
+               )
+            )
+)

.gitignore

@@ -31,6 +31,7 @@
 
 config.mk
 bli_config.h
+bli_addon.h
 
 # -- monolithic headers --
 
@@ -44,6 +45,15 @@ include/*/*.h
 
 # BLIS testsuite output file
 output.testsuite
+output.testsuite.*
 
 # BLAS test output files
 out.*
+
+# GTAGS database
+GPATH
+GRTAGS
+GTAGS
+
+# Mac DS.store files
+.DS_Store

.travis.yml

@@ -1,83 +1,132 @@
 language: c
 sudo: required
-dist: trusty
-env:
-  global:
-      secure: "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"
+dist: focal
+branches:
+  only:
+  - master
+  - dev
+  - amd
 matrix:
   include:
-  # full testsuite (all tests except for mixed datatype)
+  # full testsuite (all tests + mixed datatype (gemm_nn only) + salt + SDE + OOT)
   - os: linux
     compiler: gcc
-    env: OOT=0 TEST=1 SDE=0 THR="none" CONF="auto"
-  # mixed-datatype testsuite (gemm_nn only)
-  - os: linux
-    compiler: gcc
-    env: OOT=0 TEST=MD SDE=0 THR="none" CONF="auto"
-  # salt testsuite (fast set of operations+parameters)
-  - os: linux
-    compiler: gcc
-    env: OOT=0 TEST=SALT SDE=0 THR="none" CONF="auto"
-  # test x86_64 ukrs with SDE
-  - os: linux
-    compiler: gcc
-    env: OOT=0 TEST=0 SDE=1 THR="none" CONF="x86_64"
+    env: OOT=1 TEST=ALL SDE=1 THR="none" CONF="x86_64" \
+      PACKAGES="gcc-9 binutils"
   # openmp build
   - os: linux
     compiler: gcc
-    env: OOT=0 TEST=0 SDE=0 THR="openmp" CONF="auto"
+    env: OOT=0 TEST=FAST SDE=0 THR="openmp" CONF="auto" \
+      PACKAGES="gcc-9 binutils"
   # pthreads build
   - os: linux
     compiler: gcc
-    env: OOT=0 TEST=0 SDE=0 THR="pthreads" CONF="auto"
-  # out-of-tree build
-  - os: linux
-    compiler: gcc
-    env: OOT=1 TEST=0 SDE=0 THR="none" CONF="auto"
+    env: OOT=0 TEST=FAST SDE=0 THR="pthreads" CONF="auto" \
+      PACKAGES="gcc-9 binutils"
   # clang build
   - os: linux
     compiler: clang
-    env: OOT=0 TEST=0 SDE=0 THR="none" CONF="auto"
+    env: OOT=0 TEST=FAST SDE=0 THR="none" CONF="auto"
+      # There seems to be some difficulty installing two Clang toolchains of
+      # different versions.
+      # Use the TravisCI default.
+      # PACKAGES="clang-8 binutils"
   # macOS with system compiler (clang)
   - os: osx
     compiler: clang
-    env: OOT=0 TEST=1 SDE=0 THR="none" CONF="auto"
+    env: OOT=0 TEST=FAST SDE=0 THR="none" CONF="auto"
   # cortexa15 build and fast testsuite (qemu)
   - os: linux
     compiler: arm-linux-gnueabihf-gcc
     env: OOT=0 TEST=FAST SDE=0 THR="none" CONF="cortexa15" \
-      PACKAGES="gcc-arm-linux-gnueabihf qemu-system-arm qemu-user" \
+      CC=arm-linux-gnueabihf-gcc CXX=arm-linux-gnueabihf-g++ \
+      PACKAGES="gcc-arm-linux-gnueabihf g++-arm-linux-gnueabihf libc6-dev-armhf-cross qemu-system-arm qemu-user" \
       TESTSUITE_WRAPPER="qemu-arm -cpu cortex-a15 -L /usr/arm-linux-gnueabihf/"
   # cortexa57 build and fast testsuite (qemu)
   - os: linux
     compiler: aarch64-linux-gnu-gcc
     env: OOT=0 TEST=FAST SDE=0 THR="none" CONF="cortexa57" \
-      PACKAGES="gcc-aarch64-linux-gnu qemu-system-arm qemu-user" \
+      CC=aarch64-linux-gnu-gcc CXX=aarch64-linux-gnu-g++ \
+      PACKAGES="gcc-aarch64-linux-gnu g++-aarch64-linux-gnu libc6-dev-arm64-cross qemu-system-arm qemu-user" \
+      TESTSUITE_WRAPPER="qemu-aarch64 -L /usr/aarch64-linux-gnu/"
+  # Apple M1 (firestorm) build and fast testsuite (qemu)
+  - os: linux
+    compiler: aarch64-linux-gnu-gcc
+    env: OOT=0 TEST=FAST SDE=0 THR="none" CONF="firestorm" \
+      CC=aarch64-linux-gnu-gcc CXX=aarch64-linux-gnu-g++ \
+      PACKAGES="gcc-aarch64-linux-gnu g++-aarch64-linux-gnu libc6-dev-arm64-cross qemu-system-arm qemu-user" \
+      TESTSUITE_WRAPPER="qemu-aarch64 -L /usr/aarch64-linux-gnu/"
+  # armsve build and fast testsuite (qemu)
+  - os: linux
+    compiler: aarch64-linux-gnu-gcc-10
+    env: OOT=0 TEST=FAST SDE=0 THR="none" CONF="armsve" \
+      CC=aarch64-linux-gnu-gcc-10 CXX=aarch64-linux-gnu-g++-10 \
+      PACKAGES="gcc-10-aarch64-linux-gnu g++-10-aarch64-linux-gnu libc6-dev-arm64-cross qemu-system-arm qemu-user" \
+      TESTSUITE_WRAPPER="qemu-aarch64 -cpu max,sve=true,sve512=true -L /usr/aarch64-linux-gnu/"
+  # arm64 build and fast testsuite (qemu)
+  # NOTE: This entry omits the -cpu flag so that while both NEON and SVE kernels
+  # are compiled, only NEON kernels will be tested. (h/t to RuQing Xu)
+  - os: linux
+    compiler: aarch64-linux-gnu-gcc-10
+    env: OOT=0 TEST=FAST SDE=0 THR="none" CONF="arm64" \
+      CC=aarch64-linux-gnu-gcc-10 CXX=aarch64-linux-gnu-g++-10 \
+      PACKAGES="gcc-10-aarch64-linux-gnu g++-10-aarch64-linux-gnu libc6-dev-arm64-cross qemu-system-arm qemu-user" \
       TESTSUITE_WRAPPER="qemu-aarch64 -L /usr/aarch64-linux-gnu/"
+  # The RISC-V targets require the qemu version available in jammy or newer.
+  # When CI is upgraded, the packages should be activated and do_script.sh
+  # cleaned up.
+  # PACKAGES="qemu-user qemu-user-binfmt"
+  - os: linux
+    compiler: riscv64-unknown-linux-gcc
+    env: OOT=0 TEST=FAST SDE=0 THR="none" BLD="--disable-shared" CONF="rv64iv" \
+      CC=riscv64-unknown-linux-gnu-gcc \
+      LDFLAGS=-static
+  - os: linux
+    compiler: riscv32-unknown-linux-gcc
+    env: OOT=0 TEST=FAST SDE=0 THR="none" BLD="--disable-shared" CONF="rv32iv" \
+      CC=riscv32-unknown-linux-gnu-gcc \
+      LDFLAGS=-static
+  - os: linux
+    compiler: clang
+    env: OOT=0 TEST=FAST SDE=0 THR="none" BLD="--disable-shared" CONF="sifive_x280" \
+      CC=clang \
+      LDFLAGS=-static
 install:
-- if [ "$TRAVIS_OS_NAME" = "linux" ]; then sudo rm -f /usr/bin/as; fi
-- if [ "$TRAVIS_OS_NAME" = "linux" ]; then sudo ln -s /usr/lib/binutils-2.26/bin/as /usr/bin/as; fi
-- if [ "$TRAVIS_OS_NAME" = "linux" ]; then sudo rm -f /usr/bin/ld; fi
-- if [ "$TRAVIS_OS_NAME" = "linux" ]; then sudo ln -s /usr/lib/binutils-2.26/bin/ld /usr/bin/ld; fi
-- if [ "$CC" = "gcc" ] && [ "$TRAVIS_OS_NAME" = "linux" ]; then export CC="gcc-6"; fi
-- if [ -n "$PACKAGES" ]; then sudo apt-get install -y $PACKAGES; fi
-addons:
-  apt:
-    sources:
-    - ubuntu-toolchain-r-test
-    packages:
-    - gcc-6
-    - binutils-2.26
-    - clang
+- if [ "$CC" = "gcc"  ] && [ "$TRAVIS_OS_NAME" = "linux" ]; then export CC="gcc-9"; fi
+- if [ -n "$PACKAGES" ] && [ "$TRAVIS_OS_NAME" = "linux" ]; then sudo apt-get install -y $PACKAGES; fi
 script:
 - export DIST_PATH=.
 - pwd
 - if [ $OOT -eq 1 ]; then export DIST_PATH=`pwd`; mkdir ../oot; cd ../oot; chmod -R a-w $DIST_PATH; fi
 - pwd
-- $DIST_PATH/configure -t $THR CC=$CC $CONF
+- if [ "$CONF" = "rv64iv" ]; then
+    $DIST_PATH/travis/do_riscv.sh "$CONF";
+    export CC=$DIST_PATH/../toolchain/riscv/bin/riscv64-unknown-linux-gnu-gcc;
+    export CXX=$DIST_PATH/../toolchain/riscv/bin/riscv64-unknown-linux-gnu-g++;
+    export TESTSUITE_WRAPPER="$DIST_PATH/../toolchain/qemu-riscv64 -cpu rv64,vext_spec=v1.0,v=true,vlen=128 -B 0x100000";
+  fi
+- if [ "$CONF" = "rv32iv" ]; then
+    $DIST_PATH/travis/do_riscv.sh "$CONF";
+    export CC=$DIST_PATH/../toolchain/riscv/bin/riscv32-unknown-linux-gnu-gcc;
+    export CXX=$DIST_PATH/../toolchain/riscv/bin/riscv32-unknown-linux-gnu-g++;
+    export TESTSUITE_WRAPPER="$DIST_PATH/../toolchain/qemu-riscv32 -cpu rv32,vext_spec=v1.0,v=true,vlen=128 -B 0x100000";
+  fi
+- if [ "$CONF" = "sifive_x280" ]; then
+    $DIST_PATH/travis/do_riscv.sh "$CONF";
+    export CC=$DIST_PATH/../toolchain/riscv/bin/clang;
+    export CXX=$DIST_PATH/../toolchain/riscv/bin/clang++;
+    export TESTSUITE_WRAPPER="$DIST_PATH/../toolchain/qemu-riscv64 -cpu rv64,vext_spec=v1.0,v=true,vlen=512 -B 0x100000";
+  fi
+- $DIST_PATH/configure -p `pwd`/../install -t $THR $BLD CC=$CC $CONF
 - pwd
 - ls -l
 - $CC --version
+- $CC -v
 - make -j 2
+- make install
+- if [ "$BLD" = "" ]; then $DIST_PATH/travis/cxx/cxx-test.sh $DIST_PATH $(ls -1 include); fi
+# Qemu SVE is failing sgemmt in some cases. Skip as this issue is not observed
+# on real chip (A64fx).
+- if [ "$CONF" = "armsve" ]; then sed -i 's/.*\<gemmt\>.*/0/' $DIST_PATH/testsuite/input.operations.fast; fi
 - if [ "$TEST" != "0" ]; then travis_wait 30 $DIST_PATH/travis/do_testsuite.sh; fi
-- if [ $SDE -eq 1 ] && [ "$TRAVIS_PULL_REQUEST" = "false" ] ; then travis_wait 30 $DIST_PATH/travis/do_sde.sh; fi
+- if [ "$SDE" = "1" ]; then travis_wait 30 $DIST_PATH/travis/do_sde.sh; fi