name: Run codspeed benchmarks

on: [push, pull_request]

concurrency:
  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
  cancel-in-progress: true

permissions:
  contents: read # to fetch code (actions/checkout)

jobs:
  benchmarks:
    if: "github.repository == 'OpenMathLib/OpenBLAS'"
    strategy:
      fail-fast: false
      matrix:
        os: [ubuntu-22.04]
        fortran: [gfortran]
        build: [make]
        pyver: ["3.12"]
    runs-on: ${{ matrix.os }}
    steps:
      - uses: actions/checkout@v3
      - uses: actions/setup-python@v3
        with:
            python-version: ${{ matrix.pyver }}

      - name: Print system information
        run: |
          if [ "$RUNNER_OS" == "Linux" ]; then
            cat /proc/cpuinfo
          fi

      - name: Install Dependencies
        run: |
          if [ "$RUNNER_OS" == "Linux" ]; then
            sudo apt-get update
            sudo apt-get install -y gfortran cmake ccache libtinfo5
          else
            echo "::error::$RUNNER_OS not supported"
            exit 1
          fi

      - name: Compilation cache
        uses: actions/cache@v3
        with:
          path: ~/.ccache
          # We include the commit sha in the cache key, as new cache entries are
          # only created if there is no existing entry for the key yet.
          # GNU make and cmake call the compilers differently. It looks like
          # that causes the cache to mismatch. Keep the ccache for both build
          # tools separate to avoid polluting each other.
          key: ccache-${{ runner.os }}-${{ matrix.build }}-${{ matrix.fortran }}-${{ github.ref }}-${{ github.sha }}
          # Restore a matching ccache cache entry. Prefer same branch and same Fortran compiler.
          restore-keys: |
            ccache-${{ runner.os }}-${{ matrix.build }}-${{ matrix.fortran }}-${{ github.ref }}
            ccache-${{ runner.os }}-${{ matrix.build }}-${{ matrix.fortran }}
            ccache-${{ runner.os }}-${{ matrix.build }}

      - name: Write out the .pc
        run: |
             cd benchmark/pybench
             cat > openblas.pc << EOF
             libdir=${{ github.workspace }}
             includedir= ${{ github.workspace }}
             openblas_config= OpenBLAS 0.3.27 DYNAMIC_ARCH NO_AFFINITY Haswell MAX_THREADS=64
             version=0.0.99
             extralib=-lm -lpthread -lgfortran -lquadmath -L${{ github.workspace }} -lopenblas
             Name: openblas
             Description: OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version
             Version: ${version}
             URL: https://github.com/xianyi/OpenBLAS
             Libs: ${{ github.workspace }}/libopenblas.so -Wl,-rpath,${{ github.workspace }}
             Libs.private: -lm -lpthread -lgfortran -lquadmath -L${{ github.workspace }} -lopenblas
             Cflags: -I${{ github.workspace}}
             EOF
             cat openblas.pc

      - name: Configure ccache
        run: |
          if [ "${{ matrix.build }}" = "make" ]; then
            # Add ccache to path
            if [ "$RUNNER_OS" = "Linux" ]; then
              echo "/usr/lib/ccache" >> $GITHUB_PATH
            elif [ "$RUNNER_OS" = "macOS" ]; then
              echo "$(brew --prefix)/opt/ccache/libexec" >> $GITHUB_PATH
            else
              echo "::error::$RUNNER_OS not supported"
              exit 1
            fi
          fi
          # Limit the maximum size and switch on compression to avoid exceeding the total disk or cache quota (5 GB).
          test -d ~/.ccache || mkdir -p ~/.ccache
          echo "max_size = 300M" > ~/.ccache/ccache.conf
          echo "compression = true" >> ~/.ccache/ccache.conf
          ccache -s

      - name: Build OpenBLAS
        run: |
          case "${{ matrix.build }}" in
            "make")
              make -j$(nproc) DYNAMIC_ARCH=1 USE_OPENMP=0 FC="ccache ${{ matrix.fortran }}"
              ;;
            "cmake")
              mkdir build && cd build
              cmake -DDYNAMIC_ARCH=1 \
                    -DNOFORTRAN=0 \
                    -DBUILD_WITHOUT_LAPACK=0 \
                    -DCMAKE_VERBOSE_MAKEFILE=ON \
                    -DCMAKE_BUILD_TYPE=Release \
                    -DCMAKE_Fortran_COMPILER=${{ matrix.fortran }} \
                    -DCMAKE_C_COMPILER_LAUNCHER=ccache \
                    -DCMAKE_Fortran_COMPILER_LAUNCHER=ccache \
                    ..
              cmake --build .
              ;;
            *)
              echo "::error::Configuration not supported"
              exit 1
              ;;
          esac

      - name: Show ccache status
        continue-on-error: true
        run: ccache -s

      - name: Install benchmark dependencies
        run: pip install meson ninja numpy pytest pytest-codspeed --user

      - name: Build the wrapper
        run: |
          cd benchmark/pybench
          export PKG_CONFIG_PATH=$PWD
          meson setup build  --prefix=$PWD/build-install
          meson install -C build
          #
          # sanity check
          cd build/openblas_wrap
          python -c'import _flapack; print(dir(_flapack))'

      - name: Run benchmarks under pytest-benchmark
        run: |
          cd benchmark/pybench
          pip install pytest-benchmark
          export PYTHONPATH=$PWD/build-install/lib/python${{matrix.pyver}}/site-packages/
          OPENBLAS_NUM_THREADS=1 pytest benchmarks/bench_blas.py -k 'gesdd'

      - name: Run benchmarks
        uses: CodSpeedHQ/action@v3
        with:
          token: ${{ secrets.CODSPEED_TOKEN }}
          run: |
            cd benchmark/pybench
            export PYTHONPATH=$PWD/build-install/lib/python${{matrix.pyver}}/site-packages/
            OPENBLAS_NUM_THREADS=1 pytest benchmarks/bench_blas.py --codspeed