# NVIDIA build is separate from theRock CI due to multiple issues:
# 1. We do not want to publish any CUDA based artifacts.
# 2. CLR builds are parallel to the rest of the monorepo: we neither consume anything upstream, nor impact any downstream projects.
# Hence, the check is independent from the rest of the CI
name: HIP NVIDIA CI

on:
  pull_request:
    types:
      - opened
      - synchronize
      - reopened
  push:
    branches:
      - develop
      - release/therock-*
  workflow_dispatch:

permissions:
  contents: read

concurrency:
  group: ${{ github.workflow }}-${{ github.event.number || github.sha }}
  cancel-in-progress: true

jobs:
  nvidia-clr-build-setup:
    name: "Setup"
    runs-on: ubuntu-24.04
    env:
      BASE_REF: HEAD^
    outputs:
      should_build: ${{ steps.check_paths.outputs.should_build }}
    steps:
      - name: Checkout code
        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
        with:
          fetch-depth: 2

      - name: Check if relevant paths changed
        id: check_paths
        run: |
          SHOULD_BUILD=false

          if [ "${{ github.event_name }}" = "workflow_dispatch" ]; then
            SHOULD_BUILD=true;
          else
            CHANGED_FILES=$(git diff --name-only "${BASE_REF}")
            echo "Changed files:"
            echo "${CHANGED_FILES}"
            if echo "${CHANGED_FILES}" | grep -qE '^projects/(clr|hip|hip-tests|hipother)/|^\.github/workflows/hip-nvidia-ci\.yml'; then
              SHOULD_BUILD=true
            else
              echo "No changes in clr/hip/hip-tests/hipother — skipping NVIDIA build"
            fi
          fi
          echo "should_build=${SHOULD_BUILD}" >> "${GITHUB_OUTPUT}"

  nvidia-clr-build:
    name: "Build clr + hip-tests (NVIDIA)"
    needs: nvidia-clr-build-setup
    if: ${{ needs.nvidia-clr-build-setup.outputs.should_build == 'true' }}
    runs-on: azure-linux-scale-rocm
    container:
      # CUDA 13.2.0 devel Ubuntu 24.04
      image: nvidia/cuda@sha256:d266e59b88c295bc5fa0e4cef9064eaff84939381b09e2c3d76a5532a303e42d
    env:
      BUILD_DIR: ${{ github.workspace }}/build
    steps:
      - name: Checkout rocm-systems
        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
        with:
          path: rocm-systems
          sparse-checkout: |
            projects/clr
            projects/hip
            projects/hip-tests
            projects/hipother

      - name: Checkout llvm-project repository # we only need hipcc to build HIP on nvidia platform
        uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
        with:
          repository: "ROCm/llvm-project"
          path: "llvm-project"
          ref: 43215c73116c407735c85a180d174f718798c328 # 2026-04-14 bump in TheRock
          sparse-checkout: |
            amd/hipcc

      - name: Install system dependencies
        run: |
          apt-get update -q
          apt-get install -y --no-install-recommends git cmake ninja-build llvm-dev libnuma-dev libelf-dev libdrm-dev python3-dev python3-pip pkg-config

      - name: Build hipconfig
        run: |
          cmake -S llvm-project/amd/hipcc -B "${BUILD_DIR}/hipcc" -G Ninja -DCMAKE_BUILD_TYPE=Release
          cmake --build "${BUILD_DIR}/hipcc" --target hipconfig --parallel
          mkdir -p "${BUILD_DIR}/hipcc-install/bin"
          cp "${BUILD_DIR}/hipcc/hipconfig" "${BUILD_DIR}/hipcc-install/bin/"
          ln -s "$(which nvcc)" "${BUILD_DIR}/hipcc-install/bin/hipcc"

      - name: Configure clr (NVIDIA platform)
        run: |
          cmake -B "${BUILD_DIR}/clr" -G Ninja \
            -DCMAKE_BUILD_TYPE=Release \
            -DHIP_PLATFORM=nvidia \
            -DCLR_BUILD_HIP=ON \
            -DCLR_BUILD_OCL=OFF \
            -DHIP_COMMON_DIR="${GITHUB_WORKSPACE}/rocm-systems/projects/hip" \
            -DHIPNV_DIR="${GITHUB_WORKSPACE}/rocm-systems/projects/hipother/hipnv" \
            -DHIPCC_BIN_DIR="${BUILD_DIR}/hipcc-install/bin" \
            -DCMAKE_INSTALL_PREFIX="${BUILD_DIR}/clr-install" \
            "${GITHUB_WORKSPACE}/rocm-systems/projects/clr/"

      - name: Build clr
        run: cmake --build "${BUILD_DIR}/clr" --parallel

      - name: Install clr
        run: cmake --install "${BUILD_DIR}/clr"

      - name: Configure hip-tests (NVIDIA platform)
        run: |
          cmake -B "${BUILD_DIR}/hip-tests" -G Ninja \
            -DCMAKE_BUILD_TYPE=Release \
            -DHIP_PLATFORM=nvidia \
            -DHIPCC_BIN_DIR="${BUILD_DIR}/clr-install/bin" \
            -DCMAKE_PREFIX_PATH="${BUILD_DIR}/clr-install" \
            "${GITHUB_WORKSPACE}/rocm-systems/projects/hip-tests/catch"

      - name: Build hip-tests
        run: cmake --build "${BUILD_DIR}/hip-tests" --parallel

  nvidia-clr-build-ci-summary:
    name: "HIP NVIDIA CI Summary"
    if: always()
    needs: [nvidia-clr-build-setup, nvidia-clr-build]
    runs-on: ubuntu-24.04
    steps:
      - name: Check results
        run: |
          echo '${{ toJson(needs) }}'
          FAILED_JOBS="$(echo '${{ toJson(needs) }}' \
            | jq --raw-output \
            'map_values(select(.result != "success" and .result != "skipped")) | keys | join(",")'
          )"
          if [[ "${FAILED_JOBS}" != "" ]]; then
            echo "The following jobs failed: ${FAILED_JOBS}"
            exit 1
          fi