Configure CI on self-hosted runner

.github/workflows/self-hosted.yml

@@ -0,0 +1,112 @@
+name: Self-Hosted
+
+on:
+  push:
+    branches: [ master ]
+  pull_request:
+    branches: [ master ]
+  workflow_dispatch:
+
+env:
+  build_dir: "build"
+
+jobs:
+  build:
+    name: Build ${{ matrix.os }} GCC ${{ matrix.gcc }} CUDA ${{ matrix.cuda }}
+    runs-on: self-hosted
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - os: ubuntu-24.04
+            cuda: "12.8"
+            gcc: 13
+    env:
+      config: "Release"
+
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: recursive
+
+      - name: Set environment variables
+        run: |
+          echo "CUDA_PATH=/usr/local/cuda-12.8" >> $GITHUB_ENV
+          echo "${CUDA_PATH}/bin" >> $GITHUB_PATH
+          echo "LD_LIBRARY_PATH=${CUDA_PATH}/lib64:${LD_LIBRARY_PATH}" >> $GITHUB_ENV
+          echo "CC=/usr/bin/gcc-${{ matrix.gcc }}" >> $GITHUB_ENV
+          echo "CXX=/usr/bin/g++-${{ matrix.gcc }}" >> $GITHUB_ENV
+          echo "CUDAHOSTCXX=/usr/bin/g++-${{ matrix.gcc }}" >> $GITHUB_ENV
+          echo "CUDACXX=/usr/local/cuda-${{ matrix.cuda }}/bin/nvcc" >> $GITHUB_ENV
+
+      - name: Configure CMake build
+        run: |
+          cmake . -B ${{ env.build_dir }} -DCMAKE_BUILD_TYPE=${{ env.config }} -DCUBOOL_BUILD_TESTS=ON
+
+      - name: Build library sources
+        working-directory: ${{ env.build_dir }}
+        run: |
+          cmake --build . --target all --verbose -j `nproc`
+
+  test:
+    name: Test GPU ${{ matrix.gpu }} CUDA ${{ matrix.cuda }}
+    needs: build
+    runs-on: self-hosted
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - gpu: NVIDIA-GeForce-GT-1030
+            cuda: "12.9"
+    env:
+      unit-test-file: gpu_test_all.log
+      regression-test-file: gpu_test_regression.log
+
+    steps:
+      - name: Run unit-tests
+        working-directory: ${{ env.build_dir }}
+        run: |
+          source scripts/run_tests_all.sh | tee ${{ env.unit-test-file }}
+
+      - name: Upload unit tests resutls
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ env.unit-test-file }}
+          path: ${{ env.build_dir }}/${{ env.unit-test-file }}
+
+      - name: Check for unit tests results
+        working-directory: ${{ env.build_dir }}
+        run: |
+          if grep -q "FAILED" ${{ env.unit-test-file }}; then
+            exit 1
+          fi
+
+      - name: Run regression-tests
+        working-directory: ${{ env.build_dir }}/python
+        run: |
+          source run_tests.sh 2>&1 | tee ${{ env.regression-test-file }}
+
+      - name: Upload regression tests resutls
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ env.regression-test-file }}
+          path: ${{ env.build_dir }}/python/${{ env.regression-test-file }}
+
+      - name: Check for regression tests results
+        working-directory: ${{ env.build_dir }}/python
+        run: |
+          if grep -q "FAILED" ${{ env.regression-test-file }}; then
+            exit 1
+          fi
+
+  clean:
+    name: Cleanup workspace
+    needs: test
+    if: always()
+    runs-on: self-hosted
+
+    steps:
+      - name: Cleanup workspace
+        run: |
+          rm -rf ${{ github.workspace }}/*
+          rm -rf ${{ github.workspace }}/.*

.github/workflows/ubuntu.yml

@@ -1,5 +1,3 @@
-# Original script from https://github.com/ptheywood/cuda-cmake-github-actions
-
 name: Ubuntu
 
 on:
@@ -9,74 +7,102 @@ on:
     branches: [ master ]
   workflow_dispatch:
 
+env:
+  build_dir: "build"
+  artifact: "cubool-ubuntu-build.tar.xz"
+
 jobs:
   build:
-    name: Build ${{ matrix.os }} GCC ${{ matrix.gcc }} CUDA ${{ matrix.cuda }}
+    name: Build ${{ matrix.os }}
     runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
       matrix:
         include:
-          - os: ubuntu-18.04
-            cuda: "10.1"
-            gcc: 8
+          - os: ubuntu-22.04
     env:
-      build_dir: "build"
       config: "Release"
-      artifact: "cubool-ubuntu-build.tar.xz"
 
     steps:
-      - uses: actions/checkout@v2
+      - uses: actions/checkout@v4
         with:
           submodules: true
-      - uses: actions/setup-python@v2
-        with:
-          python-version: '3.7'
 
-      - name: Install CUDA
-        env:
-          cuda: ${{ matrix.cuda }}
+      - name: Configure CMake build
         run: |
-          source ./scripts/install_cuda_ubuntu.sh
-          if [[ $? -eq 0 ]]; then
-            # Set paths for subsequent steps, using ${CUDA_PATH}
-            echo "Adding CUDA to CUDA_PATH, PATH and LD_LIBRARY_PATH"
-            echo "CUDA_PATH=${CUDA_PATH}" >> $GITHUB_ENV
-            echo "${CUDA_PATH}/bin" >> $GITHUB_PATH
-            echo "LD_LIBRARY_PATH=${CUDA_PATH}/lib:${LD_LIBRARY_PATH}" >> $GITHUB_ENV
-          fi
-        shell: bash
+          cmake . -B ${{ env.build_dir }} -DCMAKE_BUILD_TYPE=${{ env.config }} -DCUBOOL_BUILD_TESTS=ON -DCUBOOL_WITH_CUDA=OFF
 
-      - name: Install and configure GCC and GXX
+      - name: Build library sources
+        working-directory: ${{ env.build_dir }}
         run: |
-          sudo apt-get install -y gcc-${{ matrix.gcc }} g++-${{ matrix.gcc }}
-          echo "СС=/usr/bin/gcc-${{ matrix.gcc }}" >> $GITHUB_ENV
-          echo "CXX=/usr/bin/g++-${{ matrix.gcc }}" >> $GITHUB_ENV
-          echo "CUDAHOSTCXX=/usr/bin/g++-${{ matrix.gcc }}" >> $GITHUB_ENV
+          cmake --build . --target all --verbose -j `nproc`
 
-      - name: Configure CMake build
-        run: cmake . -B ${{ env.build_dir }} -DCMAKE_BUILD_TYPE=${{ env.config }} -DCUBOOL_BUILD_TESTS=YES
+      - name: Prepare upload binary
+        run: |
+          tar cfz ${{ env.artifact }} ${{ env.build_dir }}
 
-      - name: Build library sources
+      - name: Upload binary
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ env.artifact }}
+          path: ${{ env.artifact }}
+
+  test:
+    name: Test CPU ${{ matrix.cpu }}
+    needs: build
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - os: ubuntu-22.04
+            cpu: AMD-EPYC-7763
+    env:
+      unit-test-file: cpu_test_all.log
+      regression-test-file: cpu_test_regression.log
+
+    steps:
+      - uses: actions/download-artifact@v4
+        with:
+          name: ${{ env.artifact }}
+
+      - name: Unarchive artifact
+        run: |
+          tar xzf ${{ env.artifact }}
+          rm ${{ env.artifact }}
+
+      - name: Run unit-tests
         working-directory: ${{ env.build_dir }}
-        run: cmake --build . --target all --verbose -j `nproc`
+        run: |
+          source scripts/run_tests_all.sh | tee ${{ env.unit-test-file }}
 
-      - name: Run unit-tests (sequential backend)
+      - name: Upload unit tests resutls
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ env.unit-test-file }}
+          path: ${{ env.build_dir }}/${{ env.unit-test-file }}
+
+      - name: Check for unit tests results
         working-directory: ${{ env.build_dir }}
-        run: bash scripts/run_tests_fallback.sh
-        shell: bash
+        run: |
+          if grep -q "FAILED" ${{ env.unit-test-file }}; then
+            exit 1
+          fi
 
-      - name: Run regression-tests (sequential backend)
+      - name: Run regression-tests
         working-directory: ${{ env.build_dir }}/python
-        run: bash run_tests.sh
-        shell: bash
-
-      - name: Prepare upload binary
-        shell: bash
-        run: tar cfz ${{ env.artifact }} ${{ env.build_dir }}
+        run: |
+          source run_tests.sh 2>&1 | tee ${{ env.regression-test-file }}
 
-      - name: Upload binary
-        uses: actions/upload-artifact@v2
+      - name: Upload regression tests resutls
+        uses: actions/upload-artifact@v4
         with:
-          name: ${{ env.artifact }}
-          path: ${{ env.artifact }}
+          name: ${{ env.regression-test-file }}
+          path: ${{ env.build_dir }}/python/${{ env.regression-test-file }}
+
+      - name: Check for regression tests results
+        working-directory: ${{ env.build_dir }}/python
+        run: |
+          if grep -q "FAILED" ${{ env.regression-test-file }}; then
+            exit 1
+          fi

.gitmodules

@@ -1,6 +1,3 @@
 [submodule "deps/gtest"]
 	path = deps/gtest
 	url = https://github.com/google/googletest.git
-[submodule "deps/cub"]
-	path = deps/cub
-	url = https://github.com/NVIDIA/cub.git

CMakeLists.txt

@@ -11,6 +11,8 @@ option(CUBOOL_WITH_NAIVE         "Build library with naive and naive-shared dens
 option(CUBOOL_BUILD_TESTS        "Build project unit-tests with gtest" ON)
 option(CUBOOL_COPY_TO_PY_PACKAGE "Copy compiled shared library into python package folder (for package use purposes)" ON)
 
+option(CUBOOL_USE_NSPARSE_MERGE_FUNCTOR "Use nsparse optimiztion for matrix addition function" ON)
+
 set(CUBOOL_VERSION_MAJOR 1)
 set(CUBOOL_VERSION_MINOR 0)
 set(CUBOOL_VERSION_SUB 0)
@@ -32,14 +34,6 @@ endif()
 
 # Configure cuda dependencies
 if (CUBOOL_WITH_CUDA)
-    message(STATUS "Add cub as cuda utility")
-    set(CUB_ENABLE_HEADER_TESTING OFF CACHE BOOL "" FORCE)
-    set(CUB_ENABLE_TESTING OFF CACHE BOOL "" FORCE)
-    set(CUB_ENABLE_EXAMPLES OFF CACHE BOOL "" FORCE)
-    add_subdirectory(deps/cub)
-    add_library(cub INTERFACE IMPORTED)
-    target_link_libraries(cub INTERFACE CUB::CUB)
-
     message(STATUS "Add nsparse library as crs matrix multiplication backend")
     add_subdirectory(deps/nsparse-um)
 endif()
@@ -61,4 +55,4 @@ add_subdirectory(cubool)
 file(COPY scripts DESTINATION ${CMAKE_BINARY_DIR}/)
 
 # Copy python related stuff
-file(COPY python DESTINATION ${CMAKE_BINARY_DIR}/)
+file(COPY python DESTINATION ${CMAKE_BINARY_DIR}/)

README.md

@@ -214,7 +214,7 @@ $ bash ./scripts/run_tests_all.sh
 By default, the following cmake options will be automatically enabled:
 
 - `CUBOOL_WITH_CUDA` - build library with actual cuda backend
-- `CUBOOL_WITH_SEQUENTIAL` - build library witt cpu based backend
+- `CUBOOL_WITH_SEQUENTIAL` - build library with cpu based backend
 - `CUBOOL_WITH_TESTS` - build library unit-tests collection
 
 > Note: in order to provide correct GCC version for CUDA sources compiling,

cubool/CMakeLists.txt

@@ -87,6 +87,7 @@ set(CUBOOL_C_API_SOURCES
     sources/cuBool_Matrix_Reduce2.cpp
     sources/cuBool_Matrix_EWiseAdd.cpp
     sources/cuBool_Matrix_EWiseMult.cpp
+    sources/cuBool_Matrix_EWiseMultInverted.cpp
     sources/cuBool_Vector_New.cpp
     sources/cuBool_Vector_Build.cpp
     sources/cuBool_Vector_SetElement.cpp
@@ -125,6 +126,7 @@ if (CUBOOL_WITH_CUDA)
         sources/cuda/cuda_matrix.cu
         sources/cuda/cuda_matrix_ewiseadd.cu
         sources/cuda/cuda_matrix_ewisemult.cu
+        sources/cuda/cuda_matrix_ewisemult_inverted.cu
         sources/cuda/cuda_matrix_kronecker.cu
         sources/cuda/cuda_matrix_multiply.cu
         sources/cuda/cuda_matrix_transpose.cu
@@ -147,6 +149,7 @@ if (CUBOOL_WITH_CUDA)
         sources/cuda/kernels/spgemv_t.cuh
         sources/cuda/kernels/spewiseadd.cuh
         sources/cuda/kernels/spewisemult.cuh
+        sources/cuda/kernels/spewisemultinverted.cuh
         sources/cuda/kernels/sptranspose.cuh
         sources/cuda/kernels/sptranspose2.cuh
         sources/cuda/kernels/spkron.cuh
@@ -173,6 +176,8 @@ if (CUBOOL_WITH_SEQUENTIAL)
         sources/sequential/sq_ewiseadd.hpp
         sources/sequential/sq_ewisemult.cpp
         sources/sequential/sq_ewisemult.hpp
+        sources/sequential/sq_ewisemultinverted.cpp
+        sources/sequential/sq_ewisemultinverted.hpp
         sources/sequential/sq_spgemm.cpp
         sources/sequential/sq_spgemm.hpp
         sources/sequential/sq_spgemv.cpp
@@ -201,11 +206,9 @@ target_compile_definitions(cubool PRIVATE CUBOOL_VERSION_MAJOR=${CUBOOL_VERSION_
 target_compile_definitions(cubool PRIVATE CUBOOL_VERSION_MINOR=${CUBOOL_VERSION_MINOR})
 target_compile_definitions(cubool PRIVATE CUBOOL_VERSION_SUB=${CUBOOL_VERSION_SUB})
 
-target_compile_features(cubool PUBLIC cxx_std_14)
+target_compile_definitions(cubool PRIVATE CUBOOL_USE_NSPARSE_MERGE_FUNCTOR=$<BOOL:${CUBOOL_USE_NSPARSE_MERGE_FUNCTOR}>)
 
-target_compile_options(cubool PRIVATE $<$<COMPILE_LANGUAGE:CXX>: -Wall>)
-target_compile_options(cubool PRIVATE $<$<AND:$<CONFIG:Debug>,$<COMPILE_LANGUAGE:CXX>>: -O2>)
-target_compile_options(cubool PRIVATE $<$<AND:$<CONFIG:Release>,$<COMPILE_LANGUAGE:CXX>>: -O0>)
+target_compile_features(cubool PUBLIC cxx_std_17)
 
 set_target_properties(cubool PROPERTIES CXX_STANDARD 17)
 set_target_properties(cubool PROPERTIES CXX_STANDARD_REQUIRED ON)
@@ -219,7 +222,6 @@ endforeach()
 if (CUBOOL_WITH_CUDA)
     set_target_properties(cubool PROPERTIES CUDA_STANDARD 14)
     set_target_properties(cubool PROPERTIES CUDA_STANDARD_REQUIRED ON)
-    set_target_properties(cubool PROPERTIES CUDA_SEPARABLE_COMPILATION ON)
 
     # Settings: https://arnon.dk/matching-sm-architectures-arch-and-gencode-for-various-nvidia-cards/
 
@@ -256,4 +258,4 @@ if (CUBOOL_COPY_TO_PY_PACKAGE)
         "${CMAKE_BINARY_DIR}/cubool/${LIBRARY_FILE_NAME}"
         "${CMAKE_BINARY_DIR}/python/pycubool"
         COMMENT "Copy ${LIBRARY_FILE_NAME} compiled lib into python folder")
-endif()
+endif()