ci.gpu.yml

name: CI (GPU)

on:
  push:
    branches:
      - main
      - "member-function-customization"
      - "pull-request/[0-9]+"

concurrency:
  group: ${{ github.workflow }}-on-${{ github.event_name }}-from-${{ github.ref_name }}
  cancel-in-progress: true

jobs:

  build-gpu:
    if: github.repository == 'NVIDIA/stdexec'
    name: GPU (${{ matrix.name }}, CUDA ${{ matrix.cuda }}, ${{ matrix.build }})
    strategy:
      fail-fast: false
      matrix:
        include:
          - { name: "clang 20",   cuda: "12.0", cxx: "clang++", build: "Release", tag: "llvm20-cuda12.0", gpu: "v100", sm: "70", driver: "latest", arch: "amd64" }
          - { name: "clang 20",   cuda: "12.9", cxx: "clang++", build: "Release", tag: "llvm20-cuda12.9", gpu: "v100", sm: "70", driver: "latest", arch: "amd64" }
          - { name: "nvc++ 25.7", cuda: "12.9", cxx: "mpic++",  build: "Release", tag: "nvhpc25.7",       gpu: "v100", sm: "70", driver: "latest", arch: "amd64" }
          - { name: "nvc++ 25.7", cuda: "12.9", cxx: "mpic++",  build: "Debug",   tag: "nvhpc25.7",       gpu: "v100", sm: "70", driver: "latest", arch: "amd64" }
    runs-on: linux-${{ matrix.arch }}-gpu-${{ matrix.gpu }}-${{ matrix.driver }}-1
    container:
      options: -u root
      image: rapidsai/devcontainers:25.10-cpp-${{ matrix.tag }}
      env:
        NVIDIA_VISIBLE_DEVICES: "${{ env.NVIDIA_VISIBLE_DEVICES }}"
        SCCACHE_BUCKET: "rapids-sccache-devs"
        SCCACHE_REGION: "us-east-2"
        SCCACHE_S3_KEY_PREFIX: "nvidia-stdexec-dev"
        SCCACHE_SERVER_PORT: "4225"
        SCCACHE_SERVER_LOG: "sccache=debug"
    permissions:
      id-token: write # This is required for configure-aws-credentials
      contents: read  # This is required for actions/checkout
    defaults:
      run:
        shell: su coder {0}
        working-directory: /home/coder
    steps:
      - name: Run nvidia-smi to make sure GPU is working
        run: nvidia-smi
      - name: Checkout stdexec
        uses: actions/checkout@v4
        with:
          path: stdexec
          persist-credentials: false
      - name: Setup environment
        run: |
          echo "ARTIFACT_PREFIX=${{runner.os}}-cuda${{matrix.cuda}}-${{matrix.tag}}-${{matrix.arch}}" >> "${GITHUB_ENV}"
          echo "ARTIFACT_SUFFIX=${{github.run_id}}-${{github.run_attempt}}-$RANDOM" >> "${GITHUB_ENV}"
      - id: sccache-preprocessor-cache
        name: Setup sccache preprocessor cache
        uses: actions/cache@v4
        with:
          path: /home/coder/.cache/sccache/preprocessor
          restore-keys: sccache-preprocessor-cache-${{env.ARTIFACT_PREFIX}}
          key: sccache-preprocessor-cache-${{env.ARTIFACT_PREFIX}}-${{env.ARTIFACT_SUFFIX}}
      - id: sccache-dist-toolchains-cache
        name: Setup sccache-dist client toolchains cache
        uses: actions/cache@v4
        with:
          path: /home/coder/.cache/sccache-dist-client
          restore-keys: sccache-toolchains-cache-${{env.ARTIFACT_PREFIX}}
          key: sccache-toolchains-cache-${{env.ARTIFACT_PREFIX}}-${{env.ARTIFACT_SUFFIX}}
      - if: github.repository_owner == 'NVIDIA'
        name: Get AWS credentials for sccache bucket
        uses: aws-actions/configure-aws-credentials@v4
        with:
          aws-region: us-east-2
          role-duration-seconds: 28800 # 8 hours
          role-to-assume: arn:aws:iam::279114543810:role/gha-oidc-NVIDIA
      - name: Build and test GPU schedulers
        env:
          NVCC_APPEND_FLAGS: "-t=100"
          SCCACHE_DIST_URL: "https://${{ matrix.arch }}.linux.sccache.rapids.nvidia.com"
          SCCACHE_DIST_CONNECT_TIMEOUT: "30"
          SCCACHE_DIST_FALLBACK_TO_LOCAL_COMPILE: "true"
          SCCACHE_DIST_KEEPALIVE_ENABLED: "true"
          SCCACHE_DIST_KEEPALIVE_INTERVAL: "20"
          SCCACHE_DIST_KEEPALIVE_TIMEOUT: "600"
          SCCACHE_DIST_REWRITE_INCLUDES_ONLY: ${{ matrix.cxx == 'clang++' && 'true' || 'false' }}
          SCCACHE_DIST_REQUEST_TIMEOUT: "7140"
          SCCACHE_IDLE_TIMEOUT: "7200"
        run: |
          set -e;

          if test -n "${NVHPC_ROOT:+x}"; then
            HPCX_INIT="$(find -L "$NVHPC_ROOT"/comm_libs/ -path '*/latest/hpcx-init.sh')";
            if test -n "${HPCX_INIT:+x}"; then
              . "$HPCX_INIT";
              hpcx_load;
            fi
          fi

          set -x;

          devcontainer-utils-install-sccache --repo rapidsai/sccache --version rapids;
          devcontainer-utils-init-sccache-dist                           \
              --enable-sccache-dist - <<< "                              \
              --auth-type 'token'                                        \
              --auth-token '${{ secrets.STDEXEC_BUILD_CLUSTER_SECRET }}' \
          ";

          # Copy source folder into ~/stdexec
          cp -r "${GITHUB_WORKSPACE}"/stdexec ~/;
          chown -R coder:coder ~/stdexec;
          cd ~/stdexec;

          # Configure
          cmake -S . -B build -GNinja \
            -DSTDEXEC_ENABLE_CUDA=ON \
            -DSTDEXEC_ENABLE_IO_URING=OFF \
            -DCMAKE_BUILD_TYPE=${{ matrix.build }} \
            -DCMAKE_CXX_COMPILER=${{ matrix.cxx }} \
            -DCMAKE_CUDA_COMPILER=${{ matrix.cxx }} \
            -DCMAKE_CUDA_ARCHITECTURES=${{ matrix.sm }} \
            ;

          # Compile
          cmake --build build -v -j 512;

          # Print sccache stats
          sccache -s;

          # Tests
          ctest --test-dir build --verbose --output-on-failure --timeout 300;

          # Examples
          ./build/examples/nvexec/maxwell_cpu_st --iterations=1000 --N=512 --run-cpp --run-inline-scheduler;
          ./build/examples/nvexec/maxwell_cpu_mt --iterations=1000 --N=512 --run-std --run-stdpar --run-thread-pool-scheduler;
          ./build/examples/nvexec/maxwell_gpu_s --iterations=1000 --N=512 --run-cuda --run-stdpar --run-stream-scheduler;
      - if: ${{ !cancelled() }}
        name: Upload sccache logs
        uses: actions/upload-artifact@v4
        with:
          name: sccache-client-logs-${{env.ARTIFACT_PREFIX}}-${{env.ARTIFACT_SUFFIX}}
          path: /tmp/sccache*.log
          compression-level: 9


  ci-gpu:
    runs-on: ubuntu-latest
    name: CI (GPU)
    needs:
      - build-gpu
    steps:
      - run: echo "CI (GPU) success"