gpu.yml

name: GPU Tests

on:
  push:
    branches: [ master ]
  pull_request:
    branches: [ master ]
  workflow_dispatch:

concurrency:
  group: ${{ github.workflow }}-${{ github.ref }}
  cancel-in-progress: ${{ github.ref != 'refs/heads/master'}}

jobs:
  cuda-tests:
    name: "CUDA Device Tests"
    runs-on: self-hosted

    env:
      CCACHE_DIR: ${{github.workspace}}/build/.ccache
      CCACHE_COMPRESS: true
      CCACHE_COMPRESSLEVEL: 6
      OMPI_MCA_btl_vader_single_copy_mechanism: none
      PARSEC_MCA_runtime_bind_threads: 0
      MKLROOT: /opt/intel/mkl
      Eigen3_DIR: /home/evaleev/code/install/eigen-master
      LD_LIBRARY_PATH: /home/evaleev/code/install/gcc/12/lib64

    steps:
    - uses: actions/checkout@v4

    - name: Setup ccache
      uses: actions/cache@v4
      with:
        path: ${{github.workspace}}/build/.ccache
        key: ${{ runner.os }}-cuda-ccache-${{ github.sha }}
        restore-keys: |
          ${{ runner.os }}-cuda-ccache-

    - name: Create Build Environment
      run: cmake -E make_directory ${{github.workspace}}/build

    - name: Configure CMake
      shell: bash
      working-directory: ${{github.workspace}}/build
      run: |
        cmake $GITHUB_WORKSPACE \
          -G Ninja \
          -DCMAKE_CXX_COMPILER=/home/evaleev/code/install/gcc/12/bin/g++ \
          -DCMAKE_C_COMPILER=/home/evaleev/code/install/gcc/12/bin/gcc \
          -DCMAKE_CUDA_HOST_COMPILER=/home/evaleev/code/install/gcc/12/bin/g++ \
          -DCMAKE_CUDA_COMPILER=/usr/local/cuda/bin/nvcc \
          -DCMAKE_CUDA_ARCHITECTURES=native \
          -DCUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda \
          -DCMAKE_BUILD_TYPE=Release \
          -DBUILD_SHARED_LIBS=OFF \
          -DCMAKE_INSTALL_PREFIX=${{github.workspace}}/install \
          -DTTG_EXAMPLES=ON \
          -DTTG_ENABLE_CUDA=ON \
          -DMPIEXEC_PREFLAGS='--bind-to;none;--allow-run-as-root' \
          || (cat CMakeFiles/CMakeConfigureLog.yaml && exit 1)

    - name: Build TTG
      working-directory: ${{github.workspace}}/build
      shell: bash
      run: ccache -p && ccache -z && cmake --build . && ccache -s

    - name: Build CUDA Examples
      working-directory: ${{github.workspace}}/build
      shell: bash
      run: |
        echo "Building CUDA device examples..."
        cmake --build . --target bspmm-cuda-parsec || echo "bspmm-cuda-parsec not available"
        cmake --build . --target testing_dpotrf_cuda-parsec || echo "testing_dpotrf_cuda-parsec not available"
        cmake --build . --target chain-ttg-cuda-parsec || echo "chain-ttg-cuda-parsec not available"

    - name: Install
      working-directory: ${{github.workspace}}/build
      shell: bash
      run: cmake --build . --target install

    - name: Configure Fibonacci Example
      shell: bash
      run: |
        cmake -E make_directory ${{github.workspace}}/fibonacci-build
        cd ${{github.workspace}}/fibonacci-build
        cmake ${{github.workspace}}/doc/dox/dev/devsamp/fibonacci \
          -G Ninja \
          -DCMAKE_CXX_COMPILER=/home/evaleev/code/install/gcc/12/bin/g++ \
          -DCMAKE_C_COMPILER=/home/evaleev/code/install/gcc/12/bin/gcc \
          -DCMAKE_CUDA_HOST_COMPILER=/home/evaleev/code/install/gcc/12/bin/g++ \
          -DCMAKE_CUDA_COMPILER=/usr/local/cuda/bin/nvcc \
          -DCMAKE_CUDA_ARCHITECTURES=native \
          -DCUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda \
          -DCMAKE_PREFIX_PATH="${{github.workspace}}/install" \
          || (cat CMakeFiles/CMakeConfigureLog.yaml && exit 1)

    - name: Build Fibonacci CUDA Example
      working-directory: ${{github.workspace}}/fibonacci-build
      shell: bash
      run: cmake --build . --target fibonacci_cuda-parsec

    - name: Run GPU Tests
      working-directory: ${{github.workspace}}/build
      shell: bash
      run: |
        set -e  # Exit on first error

        echo "=== GPU Device Information ==="
        nvidia-smi || echo "nvidia-smi not available"

        echo "=== Running CUDA Examples ==="

        # Test bspmm-cuda if it exists
        if [ -f "examples/bspmm-cuda-parsec" ]; then
          echo "Testing bspmm-cuda-parsec..."
          mpiexec -n 1 ./examples/bspmm-cuda-parsec
          mpiexec -n 1 ./examples/bspmm-cuda-parsec 10 10 10
        fi

        # Test testing_dpotrf_cuda if it exists
        if [ -f "examples/testing_dpotrf_cuda-parsec" ]; then
          echo "Testing testing_dpotrf_cuda-parsec..."
          mpiexec -n 1 ./examples/testing_dpotrf_cuda-parsec -N 100 -nb 25 -P 1
        fi

        # Test chain-ttg-cuda if it exists
        if [ -f "examples/chain-ttg-cuda-parsec" ]; then
          echo "Testing chain-ttg-cuda-parsec..."
          mpiexec -n 1 ./examples/chain-ttg-cuda-parsec 100
        fi

        # Test fibonacci_cuda-parsec example (built from installed TTG)
        if [ -f "${{github.workspace}}/fibonacci-build/fibonacci_cuda-parsec" ]; then
          echo "Testing fibonacci_cuda-parsec..."
          mpiexec -n 1 ${{github.workspace}}/fibonacci-build/fibonacci_cuda-parsec 20
        fi

    - name: Upload build artifacts
      if: failure()
      uses: actions/upload-artifact@v4
      with:
        name: cuda-build-logs
        path: |
          ${{github.workspace}}/build/CMakeFiles/CMakeConfigureLog.yaml
          ${{github.workspace}}/build/CMakeCache.txt