gpu-tests.yml

name: GPU Tests

on:
  # Manual trigger for GPU tests
  workflow_dispatch:
    inputs:
      backend:
        description: 'GPU backend to test'
        required: true
        default: 'all'
        type: choice
        options:
          - all
          - cuda
          - wgpu
          - metal
  # Run on PRs with GPU label
  pull_request:
    types: [labeled]

env:
  CARGO_TERM_COLOR: always
  RUST_BACKTRACE: 1

jobs:
  # CUDA GPU Tests - requires self-hosted runner with NVIDIA GPU
  cuda-tests:
    name: CUDA Tests
    if: |
      github.event_name == 'workflow_dispatch' &&
      (github.event.inputs.backend == 'all' || github.event.inputs.backend == 'cuda')
      || (github.event_name == 'pull_request' && contains(github.event.pull_request.labels.*.name, 'gpu-test'))
    runs-on: [self-hosted, gpu, cuda]
    timeout-minutes: 30
    steps:
      - uses: actions/checkout@v4

      - name: Setup Rust
        uses: dtolnay/rust-toolchain@stable

      - name: Check CUDA availability
        run: |
          nvidia-smi
          nvcc --version

      - name: Cache cargo
        uses: Swatinem/rust-cache@v2
        with:
          shared-key: "gpu-cuda"

      - name: Run CUDA codegen tests
        run: cargo test -p ringkernel-cuda-codegen --all-features

      - name: Run CUDA backend tests
        run: cargo test -p ringkernel-cuda --features cuda

      - name: Run GPU execution verification tests
        run: cargo test -p ringkernel-cuda --test gpu_execution_verify --features cuda

      - name: Run WaveSim3D GPU benchmark
        run: |
          cargo run -p ringkernel-wavesim3d --bin wavesim3d-benchmark --release --features cuda-codegen -- --quick
        continue-on-error: true

      - name: Run TxMon GPU benchmark
        run: |
          cargo run -p ringkernel-txmon --bin txmon-benchmark --release --features cuda-codegen -- --quick
        continue-on-error: true

  # WebGPU Tests - can run on any runner with Vulkan/DX12/Metal support
  wgpu-tests:
    name: WebGPU Tests
    if: |
      github.event_name == 'workflow_dispatch' &&
      (github.event.inputs.backend == 'all' || github.event.inputs.backend == 'wgpu')
      || (github.event_name == 'pull_request' && contains(github.event.pull_request.labels.*.name, 'gpu-test'))
    runs-on: [self-hosted, gpu]
    timeout-minutes: 20
    steps:
      - uses: actions/checkout@v4

      - name: Setup Rust
        uses: dtolnay/rust-toolchain@stable

      - name: Cache cargo
        uses: Swatinem/rust-cache@v2
        with:
          shared-key: "gpu-wgpu"

      - name: Run WGSL codegen tests
        run: cargo test -p ringkernel-wgpu-codegen --all-features

      - name: Run WebGPU backend tests
        run: cargo test -p ringkernel-wgpu --features wgpu-tests -- --ignored
        continue-on-error: true

  # Metal Tests - macOS only
  metal-tests:
    name: Metal Tests
    if: |
      github.event_name == 'workflow_dispatch' &&
      (github.event.inputs.backend == 'all' || github.event.inputs.backend == 'metal')
      || (github.event_name == 'pull_request' && contains(github.event.pull_request.labels.*.name, 'gpu-test'))
    runs-on: macos-latest
    timeout-minutes: 20
    steps:
      - uses: actions/checkout@v4

      - name: Setup Rust
        uses: dtolnay/rust-toolchain@stable

      - name: Cache cargo
        uses: Swatinem/rust-cache@v2
        with:
          shared-key: "gpu-metal"

      - name: Check Metal availability
        run: |
          system_profiler SPDisplaysDataType | grep -i metal || echo "Metal info not available"

      - name: Run Metal backend tests
        run: cargo test -p ringkernel-metal --features metal
        continue-on-error: true

      - name: Build Metal examples
        run: cargo build -p ringkernel --examples --features metal
        continue-on-error: true

  # CPU Backend GPU Mock Tests - runs on all platforms
  cpu-mock-tests:
    name: CPU Mock GPU Tests
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Setup Rust
        uses: dtolnay/rust-toolchain@stable

      - name: Cache cargo
        uses: Swatinem/rust-cache@v2

      - name: Run CPU backend tests (GPU mock)
        run: cargo test -p ringkernel-cpu --all-features

      - name: Run core tests with CPU backend
        run: cargo test -p ringkernel-core --all-features

      - name: Run ecosystem tests with CPU mock
        run: cargo test -p ringkernel-ecosystem --features "persistent,actix,tower,axum,grpc"

  # Performance baseline on CPU
  benchmark-baseline:
    name: Performance Baseline
    runs-on: ubuntu-latest
    if: github.event_name == 'workflow_dispatch'
    steps:
      - uses: actions/checkout@v4

      - name: Setup Rust
        uses: dtolnay/rust-toolchain@stable

      - name: Cache cargo
        uses: Swatinem/rust-cache@v2

      - name: Run CPU benchmarks
        run: cargo bench --package ringkernel -- --noplot --quick
        continue-on-error: true

      - name: Run WaveSim CPU benchmark
        run: cargo run -p ringkernel-wavesim --example benchmark --release -- --quick
        continue-on-error: true

  # Summary report
  summary:
    name: Test Summary
    needs: [cuda-tests, wgpu-tests, metal-tests, cpu-mock-tests]
    if: always()
    runs-on: ubuntu-latest
    steps:
      - name: Report Status
        run: |
          echo "## GPU Test Results" >> $GITHUB_STEP_SUMMARY
          echo "" >> $GITHUB_STEP_SUMMARY
          echo "| Backend | Status |" >> $GITHUB_STEP_SUMMARY
          echo "|---------|--------|" >> $GITHUB_STEP_SUMMARY
          echo "| CUDA | ${{ needs.cuda-tests.result }} |" >> $GITHUB_STEP_SUMMARY
          echo "| WebGPU | ${{ needs.wgpu-tests.result }} |" >> $GITHUB_STEP_SUMMARY
          echo "| Metal | ${{ needs.metal-tests.result }} |" >> $GITHUB_STEP_SUMMARY
          echo "| CPU Mock | ${{ needs.cpu-mock-tests.result }} |" >> $GITHUB_STEP_SUMMARY