TheRock/.github/workflows/multi_arch_build_windows_artifacts.yml

# Copyright Advanced Micro Devices, Inc.
# SPDX-License-Identifier: MIT

# Reusable Workflow: Build Stage Artifacts
#
# This workflow builds TheRock stage artifacts for all stages.
# Handles both generic stages (compiler-runtime, etc.) and
# per-arch stages (math-libs) using conditional logic.
#
# Artifacts flow between stages via S3 using the artifact_manager.py tool.
#
# TODO: When kpack splitting is permanently enabled, change fetch commands to use
# --generic-only instead of --amdgpu-families for efficiency. Subsequent build
# stages only need generic (host) artifacts; device-specific kpacks are not
# needed until final packaging.

name: Build Stage Artifacts

on:
  workflow_call:
    inputs:
      stage_name:
        type: string
        required: true
        description: "Stage name (e.g., compiler-runtime, math-libs)"
      stage_display_name:
        type: string
        required: true
        description: "Human-readable display name for the job"
      timeout_minutes:
        type: number
        required: true
        description: "Job timeout in minutes"
      amdgpu_family:
        type: string
        required: false
        default: ""
        description: "GPU family for per-arch stages (empty for generic stages)"
      dist_amdgpu_families:
        type: string
        required: true
        description: "Semicolon-separated list of all GPU families for dist targets"
      build_variant_cmake_preset:
        type: string
      rocm_package_version:
        type: string
        required: true
        description: "ROCm package version string"
      release_type:
        description: 'Release type; controls artifact bucket and IAM role'
        type: string
        default: ci
      repository:
        description: "Repository to checkout. Defaults to github.repository."
        type: string
        default: ""
      ref:
        description: "Branch, tag, or SHA to checkout. Defaults to the triggering ref."
        type: string
        default: ""
      build_runs_on:
        description: "Build runner label (selected by configure script with weighted distribution)"
        type: string
        default: "azure-windows-scale-rocm"
      external_repo_config:
        description: "JSON config for external repo checkout (generated by detect_external_repo_config.py)"
        type: string
        default: ""

permissions:
  contents: read

jobs:
  build_stage:
    name: ${{ inputs.stage_display_name }}
    runs-on: ${{ inputs.build_runs_on }}
    timeout-minutes: ${{ inputs.timeout_minutes }}
    permissions:
      contents: read
      id-token: write
    defaults:
      run:
        shell: bash
    env:
      STAGE_NAME: ${{ inputs.stage_name }}
      AMDGPU_FAMILIES: ${{ inputs.amdgpu_family }}
      BUILD_DIR: B:\build
      CCACHE_CONFIGPATH: ${{ github.workspace }}/.ccache/ccache.conf
      CACHE_DIR: ${{ github.workspace }}/.cache
      TEATIME_FORCE_INTERACTIVE: 0
      RELEASE_TYPE: ${{ inputs.release_type }}
      EXTERNAL_FETCH_SOURCES_ARGS: ${{ inputs.external_repo_config != '' && fromJSON(inputs.external_repo_config).fetch_sources_args || '' }}

    steps:
      # Enable git longpaths before checkout (needed for external repos with long paths)
      - name: Enable git longpaths
        if: ${{ inputs.external_repo_config != '' }}
        run: git config --system core.longpaths true

      - name: Checkout Repository
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          repository: ${{ inputs.repository || github.repository }}
          ref: ${{ inputs.ref }}

      - name: Checkout external repository
        if: ${{ inputs.external_repo_config != '' }}
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          repository: ${{ fromJSON(inputs.external_repo_config).repository }}
          ref: ${{ fromJSON(inputs.external_repo_config).ref }}
          path: ${{ fromJSON(inputs.external_repo_config).checkout_path }}

      - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7.0.0
        with:
          python-version: "3.12"

      - name: Install python deps
        run: pip install -r requirements.txt

      - name: Install requirements
        run: |
          choco source disable -n=chocolatey
          choco source add -n=internal -s http://nexus-service.nexus-ns.svc.cluster.local:8081/repository/choco-group/ --priority=1
          choco install --no-progress -y ccache
          # ninja pinned due to a bug in the 1.13.0 release:
          # https://github.com/ninja-build/ninja/issues/2616
          choco install --no-progress -y ninja --version 1.12.1
          choco install --no-progress -y strawberryperl
          echo "$PATH;C:\Strawberry\c\bin" >> $GITHUB_PATH

      - uses: iterative/setup-dvc@6c05772f52d19792ed73aedb6c2bf65ec92adebb # v3.0.0
        with:
          version: '3.62.0'

      # After other installs, so MSVC gets priority in the PATH.
      - name: Configure MSVC
        uses: ilammy/msvc-dev-cmd@0b201ec74fa43914dc39ae48a89fd1d8cb592756 # v1.13.0

      - name: Adjust git config
        run: |
          git config --global core.symlinks true
          git config --global core.longpaths true
          git config fetch.parallel 10

      - name: Runner health status
        run: |
          python build_tools/health_status.py

      - name: Setup ccache
        run: |
          python ./build_tools/setup_ccache.py \
            --release-type "${{ inputs.release_type }}" \
            --dir "$(dirname $CCACHE_CONFIGPATH)" \
            --local-path "$CACHE_DIR/ccache" \
            --log-dir "${BUILD_DIR}/logs/ccache"

      # Note: no AWS credentials should be needed here since our CI buckets
      # all currently have public read access.
      - name: Fetch inbound artifacts
        run: |
          python build_tools/artifact_manager.py fetch \
            --run-id="${{ github.run_id }}" \
            --stage="${STAGE_NAME}" \
            --amdgpu-families="${{ inputs.amdgpu_family }}" \
            --output-dir="${BUILD_DIR}" \
            --bootstrap

      - name: Fetch sources
        timeout-minutes: 30
        run: python build_tools/fetch_sources.py --stage ${STAGE_NAME} --jobs 12 --depth 1 ${EXTERNAL_FETCH_SOURCES_ARGS}

      - name: Get stage configuration
        id: stage_config
        run: |
          python build_tools/configure_stage.py \
            --stage=${STAGE_NAME} \
            --amdgpu-families="${{ inputs.amdgpu_family }}" \
            --dist-amdgpu-families="${{ inputs.dist_amdgpu_families }}" \
            --gha-output

      - name: Install stage python deps
        if: ${{ steps.stage_config.outputs.pip_install_cmd }}
        run: pip install ${{ steps.stage_config.outputs.pip_install_cmd }}

      - name: Configure
        run: |
          cmake -B "${BUILD_DIR}" -S . -GNinja \
            --preset=${{ inputs.build_variant_cmake_preset }} \
            -DTHEROCK_PACKAGE_VERSION=${{ inputs.rocm_package_version }} \
            -DCMAKE_C_COMPILER_LAUNCHER=ccache \
            -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
            ${{ inputs.external_repo_config != '' && format('-DTHEROCK_{0}_SOURCE_DIR={1}', fromJSON(inputs.external_repo_config).source_package, fromJSON(inputs.external_repo_config).checkout_path) || '' }} \
            ${{ inputs.external_repo_config != '' && fromJSON(inputs.external_repo_config).extra_cmake_options || '' }} \
            ${{ steps.stage_config.outputs.cmake_args }}

      - name: Build stage
        run: |
          cmake --build "${BUILD_DIR}" --target stage-${STAGE_NAME} therock-artifacts -- -k 0

      - name: Run build tests (Comgr, blocking)
        if: ${{ !cancelled() && inputs.stage_name == 'compiler-runtime' }}
        env:
          TEATIME_FORCE_INTERACTIVE: 1
        run: |
          cmake --build "${BUILD_DIR}" --target therock-build-tests -- -k 0

      - name: Report
        if: ${{ !cancelled() }}
        run: |
          echo "CCache Stats:"
          ccache -s -v
          echo "Artifacts:"
          ls -lh "${BUILD_DIR}"/artifacts/*.tar.xz 2>/dev/null || echo "No artifacts found"

      # TODO(scotttodd): enable once these tests pass
      #     therock-examples-cpp-sdk-user can't find hipsparselt-config.cmake
      # - name: Test Packaging
      #   run: |
      #     ctest --test-dir ${BUILD_DIR} --output-on-failure

      - name: Configure AWS credentials for artifact uploads
        if: ${{ always() }}
        uses: ./.github/actions/configure_aws_artifacts_credentials
        with:
          release_type: ${{ inputs.release_type }}

      # TODO(#3336): Attempt upload `always()`?
      #   * If some part of the build failed, partial artifacts may be useful for debugging.
      #   * This script fails if `build_dir / "artifacts"` does not exist. That's useful
      #     if the build succeeded but did not produce artifacts. That's _not_ useful
      #     if the workflow failed or was cancelled before the build even started though.
      - name: Push stage artifacts
        run: |
          python build_tools/artifact_manager.py push \
            --run-id ${{ github.run_id }} \
            --stage="${STAGE_NAME}" \
            --build-dir="${BUILD_DIR}"

      - name: Upload stage logs
        if: ${{ always() }}
        run: |
          python build_tools/github_actions/post_stage_upload.py \
            --run-id=${{ github.run_id }} \
            --stage="${STAGE_NAME}" \
            --build-dir="${BUILD_DIR}" \
            --amdgpu-family="${AMDGPU_FAMILIES}"