Multi-arch CI support

.github/workflows/therock-ci.yml

@@ -1,14 +1,8 @@
 name: TheRock CI for rocgdb
 
+# Temporarily disabled: using multi-arch CI only (therock-multi-arch-ci.yml).
+# Re-enable by restoring push/pull_request triggers.
 on:
-  push:
-    branches:
-      - amd-staging
-      - amd-staging-rocgdb-*
-  pull_request:
-    branches:
-      - amd-staging
-      - amd-staging-rocgdb-*
   workflow_dispatch:
 
 permissions:

.github/workflows/therock-multi-arch-ci-asan.yml

@@ -0,0 +1,109 @@
+# Copyright Advanced Micro Devices, Inc.
+# SPDX-License-Identifier: MIT
+
+# Multi-Arch CI Host ASAN for rocgdb
+#
+# This workflow runs host Address Sanitizer (ASAN) builds and tests
+# using TheRock's multi-arch CI pipeline.
+
+name: TheRock Multi-Arch CI ASAN
+
+on:
+  push:
+    branches:
+      - amd-staging
+      - amd-staging-rocgdb-*
+  pull_request:
+    types: [labeled]
+  workflow_dispatch:
+    inputs:
+      linux_amdgpu_families:
+        type: string
+        description: "Insert comma-separated list of Linux GPU families to build and test. ex: gfx94X, gfx120X"
+        default: ""
+      linux_test_labels:
+        type: string
+        description: "If enabled, reduce test set on Linux to the list of labels prefixed with 'test:'. ex: test:rocgdb"
+        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:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  setup:
+    if: >-
+      ${{
+        github.event_name != 'pull_request' ||
+        contains(github.event.pull_request.labels.*.name, 'ci:multi-arch-asan')
+      }}
+    uses: ROCm/TheRock/.github/workflows/setup_multi_arch.yml@97f6cb5db25304b4c74f80fa273e9f1a271c58e3 # users/lumachad/rocgdb/multi-arch-ci
+    with:
+      build_variant: "host-asan"
+      # targets gfx94X and gfx950 are selected to replicate the ASAN coverage in TheRock
+      # as ASAN builds and tests require more resources, we only target these two architectures for the time being.
+      linux_amdgpu_families: ${{ inputs.linux_amdgpu_families || 'gfx94X,gfx950' }}
+      linux_test_labels: ${{ inputs.linux_test_labels || '' }}
+      prebuilt_stages: ${{ inputs.prebuilt_stages || '' }}
+      baseline_run_id: ${{ inputs.baseline_run_id || '' }}
+      repository: ROCm/TheRock
+      ref: "97f6cb5db25304b4c74f80fa273e9f1a271c58e3" # users/lumachad/rocgdb/multi-arch-ci
+      external_repo: '{"repository":"${{ github.repository }}","ref":"${{ github.sha }}","extra_cmake_options":"-DTHEROCK_USE_EXTERNAL_ROCGDB=ON","skip_packaging":true}'
+
+  linux_build_and_test:
+    name: Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }}
+    needs: setup
+    if: >-
+      ${{
+        needs.setup.outputs.linux_build_config != '' &&
+        needs.setup.outputs.enable_build_jobs == 'true'
+      }}
+    uses: ROCm/TheRock/.github/workflows/multi_arch_ci_linux.yml@97f6cb5db25304b4c74f80fa273e9f1a271c58e3 # users/lumachad/rocgdb/multi-arch-ci
+    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 }}
+      changed_projects: "rocgdb"
+      repository: ROCm/TheRock
+      ref: "97f6cb5db25304b4c74f80fa273e9f1a271c58e3" # users/lumachad/rocgdb/multi-arch-ci
+    permissions:
+      contents: read
+      id-token: write
+
+  multi_arch_ci_summary:
+    name: Multi-Arch CI Summary
+    # always() is needed to show the job statuses regardless of failure
+    if: always()
+    needs:
+      - setup
+      - linux_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: "97f6cb5db25304b4c74f80fa273e9f1a271c58e3" # users/lumachad/rocgdb/multi-arch-ci
+          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) }}'

.github/workflows/therock-multi-arch-ci-nightly.yml

@@ -0,0 +1,37 @@
+# Copyright Advanced Micro Devices, Inc.
+# SPDX-License-Identifier: MIT
+
+# Nightly trigger for Multi-Arch CI
+#
+# This workflow triggers therock-multi-arch-ci.yml via workflow_dispatch
+# with specific GPU families. This is needed because TheRock's setup_multi_arch
+# workflow runs all families when triggered by schedule, ignoring inputs.
+# By using workflow_dispatch, TheRock respects our specified families.
+
+name: TheRock Multi-Arch CI Nightly Trigger
+
+on:
+  schedule:
+    - cron: "0 7 * * *" # Runs nightly at 7 AM UTC
+  pull_request:
+    types: [labeled]
+
+permissions:
+  actions: write
+  contents: read
+
+jobs:
+  multi-arch-ci-nightly-trigger:
+    if: >-
+      ${{
+        github.event_name != 'pull_request' ||
+        contains(github.event.pull_request.labels.*.name, 'ci:multi-arch-nightly')
+      }}
+    runs-on: ubuntu-latest
+    steps:
+      - name: Trigger Multi-Arch CI with specific GPU families
+        uses: benc-uk/workflow-dispatch@e2e5e9a103e331dad343f381a29e654aea3cf8fc # v1.2.4
+        with:
+          workflow: therock-multi-arch-ci.yml
+          ref: ${{ github.event_name == 'pull_request' && github.head_ref || 'amd-staging' }}
+          inputs: '{ "linux_amdgpu_families": "gfx94X,gfx950" }'

.github/workflows/therock-multi-arch-ci.yml

@@ -0,0 +1,112 @@
+# Copyright Advanced Micro Devices, Inc.
+# SPDX-License-Identifier: MIT
+
+# Multi-Arch CI for rocgdb
+#
+# 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
+
+name: TheRock Multi-Arch CI
+
+on:
+  pull_request:
+    types: [labeled]
+  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:rocgdb"
+        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:
+  setup:
+    if: >-
+      ${{
+        github.event_name != 'pull_request' ||
+        contains(github.event.pull_request.labels.*.name, 'ci:multi-arch')
+      }}
+    uses: ROCm/TheRock/.github/workflows/setup_multi_arch.yml@97f6cb5db25304b4c74f80fa273e9f1a271c58e3
+    with:
+      build_variant: "release"
+      linux_amdgpu_families: ${{ inputs.linux_amdgpu_families || 'gfx94X,gfx950' }}
+      linux_test_labels: ${{ inputs.linux_test_labels || '' }}
+      prebuilt_stages: ${{ inputs.prebuilt_stages || '' }}
+      baseline_run_id: ${{ inputs.baseline_run_id || '' }}
+      repository: ROCm/TheRock
+      ref: "97f6cb5db25304b4c74f80fa273e9f1a271c58e3" # users/lumachad/rocgdb/multi-arch-ci
+      external_repo: '{"repository":"${{ github.repository }}","ref":"${{ github.sha }}","extra_cmake_options":"-DTHEROCK_USE_EXTERNAL_ROCGDB=ON","skip_packaging":true}'
+
+  linux_build_and_test:
+    name: Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }}
+    needs: setup
+    if: >-
+      ${{
+        needs.setup.outputs.linux_build_config != '' &&
+        needs.setup.outputs.enable_build_jobs == 'true'
+      }}
+    uses: ROCm/TheRock/.github/workflows/multi_arch_ci_linux.yml@97f6cb5db25304b4c74f80fa273e9f1a271c58e3
+    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 }}
+      changed_projects: "rocgdb"
+      repository: ROCm/TheRock
+      ref: "97f6cb5db25304b4c74f80fa273e9f1a271c58e3" # users/lumachad/rocgdb/multi-arch-ci
+    permissions:
+      contents: read
+      id-token: write
+
+  multi_arch_ci_summary:
+    name: Multi-Arch CI Summary
+    if: always()
+    needs:
+      - setup
+      - linux_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: "97f6cb5db25304b4c74f80fa273e9f1a271c58e3" # users/lumachad/rocgdb/multi-arch-ci
+          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) }}'