rocm-systems/.github/workflows/redhat.yml

name: RedHat Linux (GCC, Python, ROCm)
run-name: redhat

on:
  push:
    branches: [ main, develop ]
    paths-ignore:
      - '*.md'
      - 'source/docs/**'
      - 'source/python/gui/**'
      - '.github/workflows/docs.yml'
      - '.github/workflows/cpack.yml'
      - '.github/workflows/containers.yml'
      - '.github/workflows/formatting.yml'
      - 'docker/**'
  pull_request:
    branches: [ main, develop ]
    paths-ignore:
      - '*.md'
      - 'source/docs/**'
      - 'source/python/gui/**'
      - '.github/workflows/docs.yml'
      - '.github/workflows/cpack.yml'
      - '.github/workflows/containers.yml'
      - '.github/workflows/formatting.yml'
      - 'docker/**'

concurrency:
  group: ${{ github.workflow }}-${{ github.ref }}
  cancel-in-progress: true

env:
  OMNITRACE_CI: ON
  OMNITRACE_TMPDIR: "%env{PWD}%/testing-tmp"

jobs:
  rhel:
    runs-on: ubuntu-20.04
    container:
      image: jrmadsen/omnitrace:ci-base-rhel-${{ matrix.os-release }}
    strategy:
      fail-fast: false
      matrix:
        compiler: ['g++']
        os-release: [ '8.7', '9.0', '9.1' ]
        rocm-version: [ '0.0', '5.3', '5.4' ]
        build-type: ['Release']
        include:
            - compiler: 'g++'
              os-release: '9.1'
              rocm-version: '5.5'
              build-type: 'Release'
            - compiler: 'g++'
              os-release: '9.1'
              rocm-version: '5.6'
              build-type: 'Release'

    steps:
    - uses: actions/checkout@v4

    - name: Configure Env
      shell: bash
      run:
        echo "CC=$(echo '${{ matrix.compiler }}' | sed 's/+/c/g')" >> $GITHUB_ENV &&
        echo "CXX=${{ matrix.compiler }}" >> $GITHUB_ENV &&
        env

    - name: Install Packages
      if: ${{ matrix.rocm-version > 0 }}
      timeout-minutes: 30
      shell: bash
      run: |
        OS_VERSION_MAJOR=$(cat /etc/os-release | grep 'VERSION_ID' | sed 's/=/ /1' | awk '{print $NF}' | sed 's/"//g' | sed 's/\./ /g' | awk '{print $1}')
        RPM_TAG=".el${OS_VERSION_MAJOR}"
        ROCM_VERSION=${{ matrix.rocm-version }}
        ROCM_MAJOR=$(echo ${ROCM_VERSION} | sed 's/\./ /g' | awk '{print $1}')
        ROCM_MINOR=$(echo ${ROCM_VERSION} | sed 's/\./ /g' | awk '{print $2}')
        ROCM_VERSN=$(( (${ROCM_MAJOR}*10000)+(${ROCM_MINOR}*100) ))
        if [ "${OS_VERSION_MAJOR}" -eq 8 ]; then PERL_REPO=powertools; else PERL_REPO=crb; fi && \
        dnf -y --enablerepo=${PERL_REPO} install perl-File-BaseDir
        yum install -y https://repo.radeon.com/amdgpu-install/${{ matrix.rocm-version }}/rhel/${{ matrix.os-release }}/amdgpu-install-${ROCM_MAJOR}.${ROCM_MINOR}.${ROCM_VERSN}-1${RPM_TAG}.noarch.rpm
        yum install -y rocm-dev rocm-smi-lib roctracer-dev rocprofiler-dev libpciaccess

    - name: Configure, Build, and Test
      timeout-minutes: 115
      shell: bash
      run:
        git config --global --add safe.directory ${PWD} &&
        cmake --version &&
        TAG="${{ github.repository_owner }}-${{ github.ref_name }}-rhel-${{ matrix.os-release }}-${{ matrix.compiler }}-python-mpip" &&
        USE_HIP=OFF &&
        if [ ${{ matrix.rocm-version }} != "0.0" ]; then USE_HIP=ON; TAG="${TAG}-rocm-${{ matrix.rocm-version }}"; fi &&
        python3 ./scripts/run-ci.py -B build
          --name ${TAG}
          --build-jobs 2
          --site GitHub
          --
          -DCMAKE_C_COMPILER=$(echo '${{ matrix.compiler }}' | sed 's/+/c/g')
          -DCMAKE_CXX_COMPILER=${{ matrix.compiler }}
          -DCMAKE_BUILD_TYPE=${{ matrix.build-type }}
          -DCMAKE_INSTALL_PREFIX=/opt/omnitrace
          -DOMNITRACE_BUILD_TESTING=ON
          -DOMNITRACE_USE_MPI=OFF
          -DOMNITRACE_USE_HIP=${USE_HIP}
          -DOMNITRACE_USE_OMPT=OFF
          -DOMNITRACE_USE_PYTHON=ON
          -DOMNITRACE_USE_MPI_HEADERS=ON
          -DOMNITRACE_CI_MPI_RUN_AS_ROOT=ON
          -DOMNITRACE_MAX_THREADS=64
          -DOMNITRACE_PYTHON_PREFIX=/opt/conda/envs
          -DOMNITRACE_PYTHON_ENVS="py3.6;py3.7;py3.8;py3.9;py3.10;py3.11"
          -DOMNITRACE_DISABLE_EXAMPLES="transpose;rccl"
          -DOMNITRACE_BUILD_NUMBER=${{ github.run_attempt }}
          --
          -LE "transpose|rccl"

    - name: Install
      timeout-minutes: 10
      run:
        cmake --build build --target install --parallel 2

    - name: Test Install
      timeout-minutes: 10
      shell: bash
      run: |
        set -v
        source /opt/omnitrace/share/omnitrace/setup-env.sh
        ./scripts/test-install.sh --test-omnitrace-{instrument,avail,sample,rewrite,runtime,python}=1

    - name: Test User API
      timeout-minutes: 10
      run: |
        set -v
        ./scripts/test-find-package.sh --install-dir /opt/omnitrace

    - name: CTest Artifacts
      if: failure()
      continue-on-error: True
      uses: actions/upload-artifact@v4
      with:
        name: ctest-${{ github.job }}-${{ strategy.job-index }}-log
        path: |
          build/*.log

    - name: Data Artifacts
      if: failure()
      continue-on-error: True
      uses: actions/upload-artifact@v4
      with:
        name: data-${{ github.job }}-${{ strategy.job-index }}-files
        path: |
          build/omnitrace-tests-config/*.cfg
          build/omnitrace-tests-output/**/*.txt
          build/omnitrace-tests-output/**/*-instr*.json

    - name: Kill Perfetto
      if: success() || failure()
      continue-on-error: True
      run: |
        set +e
        RUNNING_PROCS=$(pgrep trace_processor_shell)
        if [ -n "${RUNNING_PROCS}" ]; then kill -s 9 ${RUNNING_PROCS}; fi