TheRock/rocm-systems/.github/workflows/rocprof-trace-decoder-continuous_integration.yml

name: RocProf Trace Decoder Continuous Integration

on:
  workflow_dispatch:
  push:
    branches:
      - develop
    paths:
      - 'projects/rocprof-trace-decoder/**'
      - '!projects/rocprof-trace-decoder/*.md'
      - '!projects/rocprof-trace-decoder/CODEOWNERS'
      - '!projects/rocprof-trace-decoder/docs/**'
      - '.github/workflows/rocprof-trace-decoder-continuous_integration.yml'
      - '.github/actions/rocprofiler-sdk-util/action.yml'
      - '.github/actions/rocprofiler-sdk-util/resolve-ci-config/**'
  pull_request:
    paths:
      - 'projects/rocprof-trace-decoder/**'
      - '!projects/rocprof-trace-decoder/*.md'
      - '!projects/rocprof-trace-decoder/CODEOWNERS'
      - '!projects/rocprof-trace-decoder/docs/**'
      - '.github/workflows/rocprof-trace-decoder-continuous_integration.yml'
      - '.github/actions/rocprofiler-sdk-util/action.yml'
      - '.github/actions/rocprofiler-sdk-util/resolve-ci-config/**'

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

env:
  ROCM_PATH: "/opt/rocm"
  PYTHON_VENV_PATH: "rocprofiler-sdk"
  PYTHON_VENV_ACTIVATE: "rocprofiler-sdk/bin/activate"
  GPU_TARGETS: "gfx906 gfx908 gfx90a gfx942 gfx950 gfx1030 gfx1100 gfx1101 gfx1102 gfx1201"

  mi325_EXCLUDE_TESTS_REGEX: ""
  mi325_EXCLUDE_LABEL_REGEX: ""

  GLOBAL_CMAKE_OPTIONS: ""
  ENABLE_ROCR_BUILD: "true"
  ENABLE_HIP_CLR_BUILD: "true"
  ENABLE_ROCPROFILER_REGISTER_BUILD: "true"
  ENABLE_AQLPROFILE_BUILD: "true"
  ENABLE_ROCDECODE_BUILD: "true"
  ENABLE_ROCJPEG_BUILD: "true"

  CI_MODE: "Continuous"

  SUBMODULE_PATHS: "projects/rocprofiler-sdk projects/aqlprofile projects/rocprofiler-register projects/rocprof-trace-decoder projects/rocr-runtime projects/clr projects/hip"

jobs:
  ci_config:
    if: ${{ github.event_name != 'pull_request' || !github.event.pull_request.head.repo.fork }}
    name: Load CI config
    runs-on: ubuntu-latest
    permissions:
      contents: read
    outputs:
      gfx94x_runner: ${{ steps.runner.outputs.gfx94x_runner }}
      gfx94x_sandbox_runner: ${{ steps.runner.outputs.gfx94x_sandbox_runner }}
    steps:
      - name: Checkout workflow actions
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
          sparse-checkout: .github/actions/rocprofiler-sdk-util/resolve-ci-config
          set-safe-directory: true

      - name: Resolve runner labels
        id: runner
        uses: ./.github/actions/rocprofiler-sdk-util/resolve-ci-config

  build-and-test-ubuntu:
    if: ${{ github.event_name != 'pull_request' || !github.event.pull_request.head.repo.fork }}
    needs: ci_config
    name: Build & Test • mi325 • Ubuntu 22.04

    runs-on: ${{ needs.ci_config.outputs.gfx94x_runner }}
    container:
      image: docker.io/rocm/rocprofiler-private:ubuntu-22.04-gfx94X-latest
      credentials:
        username: ${{ secrets.ROCPROFILER_AZURE_CI_USER }}
        password: ${{ secrets.ROCPROFILER_AZURE_CI_PASS }}
      env:
        DEBIAN_FRONTEND: noninteractive
      options: --privileged

    env:
      GIT_DISCOVERY_ACROSS_FILESYSTEM: 1

    steps:
    - uses: actions/checkout@v5
      with:
        sparse-checkout: projects/rocprof-trace-decoder
        set-safe-directory: true

    - name: Install dependencies
      timeout-minutes: 10
      shell: bash
      run: |
        apt-get update
        apt-get install -y git build-essential cmake g++ python3 python3-pip python3-pyelftools libgtest-dev libgmock-dev
        git config --global --add safe.directory '*'

    - name: List Files
      shell: bash
      run: |
        echo "PATH: ${PATH}"
        which-realpath() { echo -e "\n$1 resolves to $(realpath $(which $1))"; echo "$($(which $1) --version &> /dev/stdout | head -n 1)"; }
        for i in python3 git cmake ctest gcc g++; do which-realpath $i; done
        cat ${{ env.ROCM_PATH }}/.info/version
        ls -la
        pwd

    - name: Configure
      timeout-minutes: 5
      shell: bash
      run: |
        cmake -B build -S projects/rocprof-trace-decoder \
          -DCMAKE_BUILD_TYPE=RelWithDebInfo \
          -DCMAKE_PREFIX_PATH=${{ env.ROCM_PATH }} \
          -DPython3_EXECUTABLE=$(command -v python3) \
          -DBUILD_TESTS=ON \
          -DBUILD_MARKERS=ON \
          -DBUILD_UNIT_TESTS=ON

    - name: Build
      timeout-minutes: 15
      shell: bash
      run: |
        cmake --build build -j $(nproc)

    - name: Test
      timeout-minutes: 25
      shell: bash
      run: |
        cd build
        ctest --output-on-failure --verbose

  # ---------------------------------------------------------------------------
  # Build & test rocprofiler-sdk against this trace-decoder change (RPM)
  # Mirrors the rocprofiler-sdk core-rpm job for rhel-9.5 / sles-15.6 on mi325.
  # ---------------------------------------------------------------------------
  build-and-test-rocprofiler-sdk-rpm:
    if: ${{ github.event_name != 'pull_request' || !github.event.pull_request.head.repo.fork }}
    needs: ci_config
    name: Build & Test rocprofiler-sdk • ${{ matrix.system.gpu }} • ${{ matrix.system.os }}
    timeout-minutes: 90
    strategy:
      fail-fast: false
      matrix:
        system:
          - { os: 'rhel-9.5',  gpu: 'mi325', gpu-target: 'gfx94X', build-type: 'RelWithDebInfo' }
          - { os: 'sles-15.6', gpu: 'mi325', gpu-target: 'gfx94X', build-type: 'RelWithDebInfo' }
    runs-on: ${{ needs.ci_config.outputs.gfx94x_runner }}
    container:
      image: docker.io/rocm/rocprofiler-private:${{ matrix.system.os }}-${{ matrix.system.gpu-target }}-latest
      credentials:
        username: ${{ secrets.ROCPROFILER_AZURE_CI_USER }}
        password: ${{ secrets.ROCPROFILER_AZURE_CI_PASS }}
      options:
        --privileged
        --env-file /etc/podinfo/gha-gpu-isolation-settings
    env:
      GIT_DISCOVERY_ACROSS_FILESYSTEM: 1
      OS_TYPE: ${{ matrix.system.os }}
      GPU_RUNNER: ${{ matrix.system.gpu }}

    steps:
      - name: Clone ROCProfiler SDK & AQLProfile & ROCProfiler Register & ROCR-Runtime & Trace Decoder
        uses: actions/checkout@v6
        with:
          sparse-checkout: |
            .github/actions/rocprofiler-sdk-util
            projects/rocprofiler-sdk
            projects/aqlprofile
            projects/rocprofiler-register
            projects/rocprof-trace-decoder
            projects/rocr-runtime
            projects/clr
            projects/hip
          submodules: false
          set-safe-directory: true

      - name: Compute submodule cache key
        id: submods
        shell: bash
        run: |
          git --version
          git config --global --add safe.directory '*'
          git submodule status --recursive -- ${SUBMODULE_PATHS} > .git-submodules-status-raw
          awk '{print $1,$2}' .git-submodules-status-raw > .git-submodules-status
          echo "hash=$(sha256sum .git-submodules-status | cut -d' ' -f1)" >> "$GITHUB_OUTPUT"
          # cache scoped submodule working trees plus their .git/modules gitdirs
          awk '{print $2}' .git-submodules-status-raw > .git-submodule-paths
          for p in ${SUBMODULE_PATHS}; do echo ".git/modules/${p}"; done >> .git-submodule-paths
          { echo "paths<<EOF"; cat .git-submodule-paths; echo "EOF"; } >> "$GITHUB_OUTPUT"

      - name: Restore submodule cache
        uses: actions/cache@v6
        with:
          path: |
            ${{ steps.submods.outputs.paths }}
          key: submods-scoped-v1-${{ runner.os }}-${{ steps.submods.outputs.hash }}
          restore-keys: |
            submods-scoped-v1-${{ runner.os }}-

      - name: Init/Update submodules
        run: git submodule update --init --recursive --jobs 16 -- ${SUBMODULE_PATHS}

      - name: Install requirements (venv)
        timeout-minutes: 10
        shell: bash
        working-directory: projects/rocprofiler-sdk
        run: |
          git config --global --add safe.directory '*'
          python3 -m venv ${{ env.PYTHON_VENV_PATH }}
          source ${{ env.PYTHON_VENV_ACTIVATE }}
          export PATH=/opt/rh/gcc-toolset-11/root/usr/bin:$PATH
          python3 -m pip install --upgrade pip
          python3 -m pip install --upgrade -r requirements.txt

      - name: Build ROCProfiler SDK Dependencies (installs trace-decoder under test)
        uses: ./.github/actions/rocprofiler-sdk-util
        with:
          rocm-path: ${{ env.ROCM_PATH }}
          gpu-target: ${{ matrix.system.gpu-target }}
          compiler-launcher: /usr/local/bin/sccache
          enable-rocr-build: ${{ env.ENABLE_ROCR_BUILD }}
          enable-hip-clr-build: ${{ env.ENABLE_HIP_CLR_BUILD }}
          enable-rocprofiler-register-build: ${{ env.ENABLE_ROCPROFILER_REGISTER_BUILD }}
          enable-aqlprofile-build: ${{ env.ENABLE_AQLPROFILE_BUILD }}
          enable-rocprof-trace-decoder-build: 'true'
          ccache-key: sccache-trace-decoder-${{ matrix.system.os }}-${{ needs.ci_config.outputs.gfx94x_runner }}-${{ matrix.system.gpu }}
          ccache-variant: sccache

      - name: Configure, Build, and Test
        timeout-minutes: 30
        shell: bash
        working-directory: projects/rocprofiler-sdk
        run: |
          source ${{ env.PYTHON_VENV_ACTIVATE }}
          export PATH=~/.local/bin:/opt/rh/gcc-toolset-11/root/usr/bin:$PATH
          python3 ./source/scripts/run-ci.py -B build \
            --name ${{ github.repository }}-${{ github.ref_name }}-${{ matrix.system.os }}-${{ matrix.system.gpu }}-trace-decoder-core \
            --mode ${CI_MODE} \
            --build-jobs 16 \
            --site ${{ needs.ci_config.outputs.gfx94x_runner }} \
            --gpu-targets ${{ env.GPU_TARGETS }} \
            --run-attempt ${{ github.run_attempt }} \
            -- \
            -DROCPROFILER_DEP_ROCMCORE=ON \
            -DROCPROFILER_BUILD_DOCS=OFF \
            -DCMAKE_BUILD_TYPE=${{ matrix.system.build-type }} \
            -DCMAKE_PREFIX_PATH='${{ env.ROCM_PATH }};${{ env.ROCM_PATH }}/llvm' \
            -DPython3_EXECUTABLE=$(which python3) \
            ${{ env.GLOBAL_CMAKE_OPTIONS }} \
            -- \
            --parallel 4 \
            -LE "${${{ matrix.system.gpu }}_EXCLUDE_LABEL_REGEX}" \
            -E "${${{ matrix.system.gpu }}_EXCLUDE_TESTS_REGEX}"