TheRock/rocm-systems/.github/workflows/therock-multi-arch-ci.yml

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

# Multi-Arch CI for rocm-systems
#
# This workflow replicates TheRock's multi-arch CI pipeline by calling
# TheRock's reusable workflows directly. It provides sharded, multi-stage
# builds with per-architecture testing.
#
# Key differences from therock-ci.yml:
# - Uses TheRock's multi-arch sharded build pipeline (foundation -> compiler-runtime -> math-libs)
# - Supports prebuilt stages to skip certain build stages
# - Better parallelization across GPU architectures

# TheRock ref: using main until full migration, then switch to pinned SHA (TheRock#3343)
# NOTE: update all `@<ref>` refs and `ref:` inputs together when switching

# =============================================================================
# TEMPORARY: MI455 (gfx125X) bringup configuration
# TODO(geomin12): Remove this block once MI455 is fully enabled in therock-ci-config
#
# The gfx125x_family_overrides JSON configures MI455-specific test settings:
# - test-runs-on: Custom runner label for MI455 hardware
# - test_labels_for_family: Limit tests to hip-tests, rocrtst (sanity always runs)
# - Container options with ulimit for RCCL compatibility
# =============================================================================

name: TheRock Multi-Arch CI

on:
  workflow_dispatch:
    inputs:
      linux_amdgpu_families:
        type: string
        description: "Insert comma-separated list of Linux GPU families to build and test. ex: gfx94X, gfx120X (empty = skip linux)"
        default: ""
      linux_test_labels:
        type: string
        description: "If enabled, reduce test set on Linux to the list of labels prefixed with 'test:'. ex: test:rocprim, test:hipcub"
        default: ""
      windows_amdgpu_families:
        type: string
        description: "Insert comma-separated list of Windows GPU families to build and test. ex: gfx94X, gfx120X (empty = skip windows)"
        default: ""
      windows_test_labels:
        type: string
        description: "If enabled, reduce test set on Windows to the list of labels prefixed with 'test:' ex: test:rocprim, test:hipcub"
        default: ""
      prebuilt_stages:
        type: string
        default: ""
        description: "Comma-separated build stages to skip (or 'all' for all stages); artifacts are copied from baseline_run_id instead"
      baseline_run_id:
        type: string
        default: ""
        description: "Workflow run ID to copy prebuilt stage artifacts from; required when prebuilt_stages is set"

permissions:
  contents: read

concurrency:
  # A PR number if a pull request and otherwise the commit hash. This cancels
  # queued and in-progress runs for the same PR (presubmit) or commit
  # (postsubmit). The workflow name is prepended to avoid conflicts between
  # different workflows.
  group: ${{ github.workflow }}-${{ github.event.number || github.sha }}
  cancel-in-progress: true

jobs:
  configure:
    name: Configure Changed Projects
    runs-on: ubuntu-24.04
    outputs:
      changed_projects: ${{ steps.configure.outputs.changed_projects }}
      run_all_tests: ${{ steps.configure.outputs.run_all_tests }}
      skip_tests: ${{ steps.configure.outputs.skip_tests }}
    steps:
      - name: Checkout repository config
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
          sparse-checkout: .github/repos-config.json
          sparse-checkout-cone-mode: false

      - name: Checkout TheRock for configure script
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
          repository: ROCm/TheRock
          ref: main
          path: _therock
          sparse-checkout: build_tools/github_actions
          sparse-checkout-cone-mode: true

      - name: Configure CI
        id: configure
        run: |
          python3 _therock/build_tools/github_actions/configure_external_repo_ci.py \
            --event-name "${{ github.event_name }}" \
            --github-repo "${{ github.repository }}" \
            --base-sha "${{ github.event.pull_request.base.sha }}" \
            --head-sha "${{ github.event.pull_request.head.sha }}" \
            --config-path ".github/repos-config.json"

  setup:
    needs: configure
    uses: ROCm/TheRock/.github/workflows/setup_multi_arch.yml@main
    with:
      build_variant: "release"
      # Limit GPU families for rocm-systems CI
      # Note: gfx125X is added for MI455 bringup with limited tests (hip-tests, rocrtst, sanity)
      linux_amdgpu_families: ${{ inputs.linux_amdgpu_families || 'gfx94X,gfx950,gfx125X' }}
      windows_amdgpu_families: ${{ inputs.windows_amdgpu_families || 'gfx1151' }}
      linux_test_labels: ${{ inputs.linux_test_labels || '' }}
      windows_test_labels: ${{ inputs.windows_test_labels || '' }}
      prebuilt_stages: ${{ inputs.prebuilt_stages || '' }}
      baseline_run_id: ${{ inputs.baseline_run_id || '' }}
      repository: ROCm/TheRock
      ref: main
      # TEMPORARY: MI455 bringup - pass family_overrides to configure test runner and limited tests
      # TODO(geomin12): Remove family_overrides once MI455 is in therock-ci-config
      # NOTE: MI455 machine is down - test-runs-on set to "" until back online (was: linux-mi455-gpu-rocm)
      external_repo: >-
        {
          "repository": "${{ github.repository }}",
          "ref": "${{ github.sha }}",
          "family_overrides": {
            "gfx125x": {
              "linux": {
                "test-runs-on": "",
                "test_labels_for_family": ["test:hip-tests", "test:rocrtst"],
                "fetch-gfx-targets": ["gfx1250"]
              }
            }
          }
        }

  linux_build_and_test:
    name: Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }}
    needs: [configure, setup]
    if: >-
      ${{
        needs.setup.outputs.linux_build_config != '' &&
        needs.setup.outputs.enable_build_jobs == 'true' &&
        needs.configure.outputs.skip_tests != 'true'
      }}
    uses: ROCm/TheRock/.github/workflows/multi_arch_ci_linux.yml@main
    secrets: inherit
    with:
      build_config: ${{ needs.setup.outputs.linux_build_config }}
      test_labels: ${{ needs.setup.outputs.linux_test_labels }}
      rocm_package_version: ${{ needs.setup.outputs.rocm_package_version }}
      test_type: ${{ needs.setup.outputs.test_type }}
      external_repo_config: ${{ needs.setup.outputs.external_repo_config }}
      # Empty string when run_all_tests=true triggers full test run in downstream workflow
      changed_projects: ${{ needs.configure.outputs.run_all_tests == 'true' && '' || needs.configure.outputs.changed_projects }}
      repository: ROCm/TheRock
      ref: main
    permissions:
      contents: read
      id-token: write

  windows_build_and_test:
    name: Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }}
    needs: [configure, setup]
    if: >-
      ${{
        needs.setup.outputs.windows_build_config != '' &&
        needs.setup.outputs.enable_build_jobs == 'true' &&
        needs.configure.outputs.skip_tests != 'true'
      }}
    uses: ROCm/TheRock/.github/workflows/multi_arch_ci_windows.yml@main
    secrets: inherit
    with:
      build_config: ${{ needs.setup.outputs.windows_build_config }}
      test_labels: ${{ needs.setup.outputs.windows_test_labels }}
      rocm_package_version: ${{ needs.setup.outputs.rocm_package_version }}
      test_type: ${{ needs.setup.outputs.test_type }}
      external_repo_config: ${{ needs.setup.outputs.external_repo_config }}
      # Empty string when run_all_tests=true triggers full test run in downstream workflow
      changed_projects: ${{ needs.configure.outputs.run_all_tests == 'true' && '' || needs.configure.outputs.changed_projects }}
      repository: ROCm/TheRock
      ref: main
    permissions:
      contents: read
      id-token: write

  multi_arch_ci_summary:
    name: Multi-Arch CI Summary
    if: always()
    needs:
      - setup
      - linux_build_and_test
      - windows_build_and_test
    runs-on: ubuntu-24.04
    steps:
      - name: Checkout TheRock repository
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
          repository: "ROCm/TheRock"
          ref: main
          sparse-checkout: build_tools/github_actions
          sparse-checkout-cone-mode: true

      - name: Evaluate workflow results
        env:
          GITHUB_TOKEN: ${{ github.token }}
        run: |
          python build_tools/github_actions/workflow_summary.py \
            --needs-json '${{ toJSON(needs) }}'