Merge branch 'ver0.9.0' into main

jrenaud90 · May 23, 2024 · 02c0c57 · 02c0c57
2 parents da53ff3 + fad336d
commit 02c0c57
Show file tree

Hide file tree

Showing 26 changed files with 624 additions and 707 deletions.
diff --git a/.github/workflows/build_wheels_main_pypi.yml b/.github/workflows/build_wheels_main_pypi.yml
@@ -23,8 +23,8 @@ jobs:
         path: dist/*.tar.gz
         name: artifact_sdist        
 
-  build_macos_wheel:
-    name: Build MacOS wheel for Python ${{ matrix.python }}
+  build_macos_latest_wheel:
+    name: Build MacOS (latest) wheel for Python ${{ matrix.python }}
     runs-on: macos-latest
     strategy:
       # Ensure that a wheel builder finishes even if another fails
@@ -69,7 +69,56 @@ jobs:
       - uses: actions/upload-artifact@v4
         with:
           path: ./wheelhouse/*.whl
-          name: artifact_Macos_${{ matrix.python }}_whl
+          name: artifact_MacosLatest_${{ matrix.python }}_whl
+
+  build_macos_13_wheel:
+    name: Build MacOS-13 (x64-86) wheel for Python ${{ matrix.python }}
+    runs-on: macos-13
+    strategy:
+      # Ensure that a wheel builder finishes even if another fails
+      fail-fast: false
+      matrix:
+        python:
+          - "3.8"
+          - "3.9"
+          - "3.10"
+          - "3.11"
+          - "3.12"
+    steps:
+      - name: Checkout CyRK
+        uses: actions/checkout@v4
+
+      - name: Install Anaconda
+        uses: conda-incubator/setup-miniconda@v3
+        with:
+          activate-environment: cyrk_test
+          auto-update-conda: true
+          python-version: ${{ matrix.python }}
+          auto-activate-base: false
+
+      - name: Conda info
+        shell: bash -el {0}
+        run: conda info
+
+      - name: install gcc-13  # openMP does not play nice with clang on MacOS; also some versions of macos use older gcc so there is a need to install latest.
+        shell: bash -el {0}
+        run: |
+          brew install gcc@13
+
+      - name: Install package
+        shell: bash -el {0}
+        run: |
+          conda install nomkl
+          conda install pytest matplotlib numba numpy scipy cython
+          python -m pip install --upgrade build
+          export CC='gcc-13'
+          export CPPFLAGS="$CPPFLAGS -Xpreprocessor -fopenmp -lgomp"
+          python -m build --wheel --outdir ./wheelhouse/
+
+      - uses: actions/upload-artifact@v4
+        with:
+          path: ./wheelhouse/*.whl
+          name: artifact_Macos13_${{ matrix.python }}_whl
 
   build_other_wheels:
     name: Build ${{ matrix.platform }} wheels
@@ -97,7 +146,7 @@ jobs:
 
   check_dist:
     name: Check sdist
-    needs: [make_sdist, build_macos_wheel, build_other_wheels]
+    needs: [make_sdist, build_macos_latest_wheel, build_macos_13_wheel, build_other_wheels]
     runs-on: ubuntu-22.04
     steps:
       - uses: actions/download-artifact@v4
@@ -109,7 +158,7 @@ jobs:
 
   upload_all:
     name: Upload wheels to PyPI
-    needs: [build_macos_wheel, build_other_wheels, make_sdist, check_dist]
+    needs: [build_macos_latest_wheel, build_macos_13_wheel, build_other_wheels, make_sdist, check_dist]
     runs-on: ubuntu-latest
     steps:
       - uses: actions/download-artifact@v4

diff --git a/.github/workflows/push_tests_mac.yml b/.github/workflows/push_tests_mac.yml
@@ -7,11 +7,11 @@ on: [push, pull_request, workflow_dispatch]
 
 jobs:
 
-  test-macos:
+  test-macos-latest:
     defaults:
       run:
         shell: bash -el {0}
-    name: Test CyRK on MacOS
+    name: Test CyRK on MacOS (latest; arm64)
     runs-on: macos-latest
     strategy:
       fail-fast: false
@@ -51,3 +51,49 @@ jobs:
       - name: Run pytest
         shell: bash -el {0}
         run: pytest --capture=sys -v Tests/
+
+  test-macos-13:
+    defaults:
+      run:
+        shell: bash -el {0}
+    name: Test CyRK on MacOS (v13; x64-86)
+    runs-on: macos-13
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version:
+          - "3.8"
+          - "3.9"
+          - "3.10"
+          - "3.11"
+          - "3.12"
+    steps:
+      - uses: actions/checkout@v4
+      - uses: conda-incubator/setup-miniconda@v3
+        with:
+          activate-environment: cyrk_test
+          auto-update-conda: true
+          python-version: ${{ matrix.python-version }}
+          auto-activate-base: false
+
+      - name: Conda info
+        shell: bash -el {0}
+        run: conda info
+
+      - name: install gcc-13 # openMP does not play nice with clang on MacOS; also some versions of macos use older gcc so there is a need to install latest.
+        shell: bash -el {0}
+        run: |
+          brew install gcc@13
+
+      - name: Install package
+        shell: bash -el {0}
+        run: |
+          conda install nomkl
+          conda install pytest matplotlib numba numpy scipy cython
+          export CC="gcc-13"
+          export CPPFLAGS="$CPPFLAGS -Xpreprocessor -fopenmp -lgomp"
+          python -m pip install -v .
+        
+      - name: Run pytest
+        shell: bash -el {0}
+        run: pytest --capture=sys -v Tests/
diff --git a/Benchmarks/CyRK - SciPy Comparison.ipynb b/Benchmarks/CyRK - SciPy Comparison.ipynb
diff --git a/Benchmarks/CyRK_CySolver.pdf b/Benchmarks/CyRK_CySolver.pdf
diff --git a/Benchmarks/CyRK_SciPy_Compare_predprey_v0-9-0.png b/Benchmarks/CyRK_SciPy_Compare_predprey_v0-9-0.png
diff --git a/Benchmarks/CyRK_cyrk_ode.pdf b/Benchmarks/CyRK_cyrk_ode.pdf
diff --git a/Benchmarks/CyRK_numba.pdf b/Benchmarks/CyRK_numba.pdf
diff --git a/Benchmarks/SciPy.pdf b/Benchmarks/SciPy.pdf
diff --git a/...ks/CyRK_SciPy_Compare_predprey_v0-8-4.png → ...ve/CyRK_SciPy_Compare_predprey_v0-8-4.png b/...ks/CyRK_SciPy_Compare_predprey_v0-8-4.png → ...ve/CyRK_SciPy_Compare_predprey_v0-8-4.png
diff --git a/...ks/CyRK_SciPy_Compare_predprey_v0-8-6.png → ...ve/CyRK_SciPy_Compare_predprey_v0-8-6.png b/...ks/CyRK_SciPy_Compare_predprey_v0-8-6.png → ...ve/CyRK_SciPy_Compare_predprey_v0-8-6.png
diff --git a/CHANGES.md b/CHANGES.md
@@ -2,6 +2,17 @@
 
 ## 2024
 
+### v0.9.0 (2024-05-22)
+
+Major Changes:
+- Shifted from using the Python-based `PyMem_Alloc`, `PyMem_Free` to c-based `malloc`, `free`. 
+- CySolver `_solve` method is now gil-free.
+  - This has led to a 35%--230% speed boost at low values of steps (faster start up).
+
+Other Changes:
+- CI will now build x64-86 and arm64 wheels for MacOS (change suggested by @dihm in [#49](https://github.com/jrenaud90/CyRK/issues/49)).
+  - Did have to use this `nomkl` [workaround](https://stackoverflow.com/questions/53014306/error-15-initializing-libiomp5-dylib-but-found-libiomp5-dylib-already-initial) which may cause problems. TBD.
+
 #### v0.8.8 (2024-04-30)
 
 New Features:

diff --git a/CyRK/cy/common.pxd b/CyRK/cy/common.pxd
@@ -32,7 +32,7 @@ cdef void interpolate(
         size_t t_len_reduced,
         size_t target_len,
         bint is_complex
-        ) noexcept
+        ) noexcept nogil
 
 cdef size_t find_expected_size(
         size_t y_size,

diff --git a/CyRK/cy/common.pyx b/CyRK/cy/common.pyx
@@ -6,6 +6,7 @@ from libc.math cimport fmax, fmin, floor
 from libc.math cimport INFINITY as INF
 from libc.float cimport DBL_EPSILON as EPS
 from libc.stdint cimport SIZE_MAX, INT32_MAX
+from libc.stdlib cimport free
 
 from CyRK.utils.utils cimport allocate_mem, reallocate_mem, free_mem
 from CyRK.array.interp cimport interp_array_ptr, interp_complex_array_ptr
@@ -46,7 +47,7 @@ cdef void interpolate(
         size_t t_len_reduced,
         size_t target_len,
         bint is_complex
-        ) noexcept:
+        ) noexcept nogil:
     """ Interpolate the results of a successful integration over the user provided time domain, `time_domain_full`. """
 
     # Setup loop variables
@@ -95,10 +96,10 @@ cdef void interpolate(
     finally:
         # Release memory of any temporary variables
         if not (array_slice_ptr is NULL):
-            free_mem(array_slice_ptr)
+            free(array_slice_ptr)
             array_slice_ptr = NULL
         if not (interpolated_array_slice_ptr is NULL):
-            free_mem(interpolated_array_slice_ptr)
+            free(interpolated_array_slice_ptr)
             interpolated_array_slice_ptr = NULL
 
 cdef size_t find_expected_size(

diff --git a/CyRK/cy/cyrk.pyx b/CyRK/cy/cyrk.pyx
@@ -3,6 +3,8 @@
 
 import cython
 
+from libc.stdlib cimport free
+
 import numpy as np
 cimport numpy as np
 
@@ -867,14 +869,14 @@ def cyrk_ode(
         else:
             # Clear the storage arrays used during the step loop
             if not (time_domain_array_ptr is NULL):
-                free_mem(time_domain_array_ptr)
+                free(time_domain_array_ptr)
                 time_domain_array_ptr = NULL
             if not (y_results_array_ptr is NULL):
-                free_mem(y_results_array_ptr)
+                free(y_results_array_ptr)
                 y_results_array_ptr = NULL
             if capture_extra:
                 if not (extra_array_ptr is NULL):
-                    free_mem(extra_array_ptr)
+                    free(extra_array_ptr)
                     extra_array_ptr = NULL
 
             # Integration was not successful. Leave the solution pointers as length 1 nan arrays.
@@ -965,17 +967,17 @@ def cyrk_ode(
 
                 # Replace old pointers with new interpolated pointers and release the memory for the old stuff
                 if not (solution_extra_ptr is NULL):
-                    free_mem(solution_extra_ptr)
+                    free(solution_extra_ptr)
                 solution_extra_ptr = interpolated_solution_extra_ptr
                 interpolated_solution_extra_ptr = NULL
 
             # Replace old pointers with new interpolated pointers and release the memory for the old stuff
             if not (solution_t_ptr is NULL):
-                free_mem(solution_t_ptr)
+                free(solution_t_ptr)
             solution_t_ptr = interpolated_solution_t_ptr
             interpolated_solution_t_ptr = NULL
             if not (solution_y_ptr is NULL):
-                free_mem(solution_y_ptr)
+                free(solution_y_ptr)
             solution_y_ptr = interpolated_solution_y_ptr
             interpolated_solution_y_ptr = NULL
 
@@ -998,14 +1000,14 @@ def cyrk_ode(
                     solution_y_view[extra_start + j, i] = solution_extra_ptr[i * num_extra + j]
         # Free solution arrays
         if not (solution_t_ptr is NULL):
-            free_mem(solution_t_ptr)
+            free(solution_t_ptr)
             solution_t_ptr = NULL
         if not (solution_y_ptr is NULL):
-            free_mem(solution_y_ptr)
+            free(solution_y_ptr)
             solution_y_ptr = NULL
         if capture_extra:
             if not (solution_extra_ptr is NULL):
-                free_mem(solution_extra_ptr)
+                free(solution_extra_ptr)
                 solution_extra_ptr = NULL
 
         # Update integration message
@@ -1027,55 +1029,55 @@ def cyrk_ode(
     finally:
         # Free pointers made from user inputs
         if not (tol_ptrs is NULL):
-            free_mem(tol_ptrs)
+            free(tol_ptrs)
             tol_ptrs = NULL
         if not (t_eval_ptr is NULL):
-            free_mem(t_eval_ptr)
+            free(t_eval_ptr)
             t_eval_ptr = NULL
 
         # Free pointers used to track y, dydt, and any extra outputs
         if not (y_storage_ptrs is NULL):
-            free_mem(y_storage_ptrs)
+            free(y_storage_ptrs)
             y_storage_ptrs = NULL
         if not (extra_output_init_ptr is NULL):
-            free_mem(extra_output_init_ptr)
+            free(extra_output_init_ptr)
             extra_output_init_ptr = NULL
         if not (extra_output_ptr is NULL):
-            free_mem(extra_output_ptr)
+            free(extra_output_ptr)
             extra_output_ptr = NULL
 
         # Free RK Temp Storage Array
         if not (K_ptr is NULL):
-            free_mem(K_ptr)
+            free(K_ptr)
             K_ptr = NULL
 
         # Free other pointers that should have been freed in main loop, but in case of an exception they were missed.
         if not (solution_t_ptr is NULL):
-            free_mem(solution_t_ptr)
+            free(solution_t_ptr)
             solution_t_ptr = NULL
         if not (solution_y_ptr is NULL):
-            free_mem(solution_y_ptr)
+            free(solution_y_ptr)
             solution_y_ptr = NULL
         if not (solution_extra_ptr is NULL):
-            free_mem(solution_extra_ptr)
+            free(solution_extra_ptr)
             solution_extra_ptr = NULL
         if not (interpolated_solution_t_ptr is NULL):
-            free_mem(interpolated_solution_t_ptr)
+            free(interpolated_solution_t_ptr)
             interpolated_solution_t_ptr= NULL
         if not (interpolated_solution_y_ptr is NULL):
-            free_mem(interpolated_solution_y_ptr)
+            free(interpolated_solution_y_ptr)
             interpolated_solution_y_ptr = NULL
         if not (interpolated_solution_extra_ptr is NULL):
-            free_mem(interpolated_solution_extra_ptr)
+            free(interpolated_solution_extra_ptr)
             interpolated_solution_extra_ptr = NULL
         if not (time_domain_array_ptr is NULL):
-            free_mem(time_domain_array_ptr)
+            free(time_domain_array_ptr)
             time_domain_array_ptr = NULL
         if not (y_results_array_ptr is NULL):
-            free_mem(y_results_array_ptr)
+            free(y_results_array_ptr)
             y_results_array_ptr = NULL
         if not (extra_array_ptr is NULL):
-            free_mem(extra_array_ptr)
+            free(extra_array_ptr)
             extra_array_ptr = NULL
 
     return solution_t, solution_y, success, message