test.yml

name: Build and Test

on:
  push:
    branches: [main, metal-backend]
  pull_request:
    branches: [main, metal-backend]

env:
  BUILD_TYPE: Release

jobs:
  # Linux CPU-only build and tests
  linux-cpu:
    runs-on: ubuntu-latest

    steps:
    - uses: actions/checkout@v4

    - name: Cache CMake FetchContent dependencies
      uses: actions/cache@v4
      with:
        path: build/_deps
        key: deps-${{ runner.os }}-${{ runner.arch }}-${{ hashFiles('CMakeLists.txt') }}
        restore-keys: |
          deps-${{ runner.os }}-${{ runner.arch }}-

    - name: Install dependencies
      run: |
        sudo apt-get update
        sudo apt-get install -y libopenblas-dev cmake build-essential

    - name: Configure CMake
      run: |
        cmake -S . -B build \
          -DCMAKE_BUILD_TYPE=${{ env.BUILD_TYPE }} \
          -DSPRUX_USE_CUBLAS=OFF \
          -DSPRUX_USE_METAL=OFF \
          -DSPRUX_USE_OPENCL=OFF \
          -DSPRUX_BUILD_TESTS=ON \
          -DSPRUX_BUILD_EXAMPLES=ON

    - name: Build
      run: cmake --build build --config ${{ env.BUILD_TYPE }} -j"$(nproc)"

    - name: Run Tests
      run: ctest --test-dir build --output-on-failure -j"$(nproc)"

    - name: Scipy cross-validation (CPU)
      run: |
        python3 -m venv .venv
        .venv/bin/pip install --quiet scipy numpy
        .venv/bin/python tests/scipy_ring_comparison.py --build-dir build --test-binary SequenceSolveTest

  # Linux with OpenCL (CPU backend via pocl)
  linux-opencl:
    runs-on: ubuntu-latest

    steps:
    - uses: actions/checkout@v4

    - name: Cache CMake FetchContent dependencies
      uses: actions/cache@v4
      with:
        path: build/_deps
        key: deps-${{ runner.os }}-${{ runner.arch }}-${{ hashFiles('CMakeLists.txt') }}
        restore-keys: |
          deps-${{ runner.os }}-${{ runner.arch }}-

    - name: Install dependencies
      run: |
        sudo apt-get update
        sudo apt-get install -y libopenblas-dev cmake build-essential \
          opencl-headers ocl-icd-opencl-dev pocl-opencl-icd libclblast-dev

    - name: Configure CMake
      run: |
        cmake -S . -B build \
          -DCMAKE_BUILD_TYPE=${{ env.BUILD_TYPE }} \
          -DSPRUX_USE_CUBLAS=OFF \
          -DSPRUX_USE_METAL=OFF \
          -DSPRUX_USE_OPENCL=ON \
          -DSPRUX_BUILD_TESTS=ON \
          -DSPRUX_BUILD_EXAMPLES=OFF

    - name: Build
      run: cmake --build build --config ${{ env.BUILD_TYPE }} -j"$(nproc)"

    - name: Run Tests (OpenCL via PoCL CPU backend)
      run: ctest --test-dir build --output-on-failure -j"$(nproc)"

  # macOS with Metal GPU tests
  macos-metal:
    runs-on: macos-latest-xlarge  # Bare-metal Apple Silicon with GPU access

    steps:
    - uses: actions/checkout@v4

    - name: Cache CMake FetchContent dependencies
      uses: actions/cache@v4
      with:
        path: build/_deps
        key: deps-${{ runner.os }}-${{ runner.arch }}-${{ hashFiles('CMakeLists.txt') }}
        restore-keys: |
          deps-${{ runner.os }}-${{ runner.arch }}-

    - name: Install dependencies
      run: |
        brew install openblas llvm

    - name: Configure CMake
      run: |
        LLVM_PREFIX="$(brew --prefix llvm)"
        OPENBLAS_PREFIX="$(brew --prefix openblas)"
        export PATH="${LLVM_PREFIX}/bin:$PATH"
        export LDFLAGS="-L${OPENBLAS_PREFIX}/lib"
        export CPPFLAGS="-I${OPENBLAS_PREFIX}/include"
        cmake -S . -B build \
          -DCMAKE_BUILD_TYPE=${{ env.BUILD_TYPE }} \
          -DSPRUX_USE_CUBLAS=OFF \
          -DSPRUX_USE_METAL=ON \
          -DSPRUX_USE_OPENCL=OFF \
          -DSPRUX_BUILD_TESTS=ON \
          -DSPRUX_BUILD_EXAMPLES=ON \
          -DBLA_VENDOR=OpenBLAS \
          -DCMAKE_PREFIX_PATH="${OPENBLAS_PREFIX}"

    - name: Build
      run: |
        LLVM_PREFIX="$(brew --prefix llvm)"
        export PATH="${LLVM_PREFIX}/bin:$PATH"
        cmake --build build --config ${{ env.BUILD_TYPE }} -j"$(sysctl -n hw.ncpu)"

    - name: Run all tests (CPU + Metal GPU)
      run: ctest --test-dir build --output-on-failure -j"$(sysctl -n hw.ncpu)"

    - name: Scipy cross-validation (CPU + Metal)
      run: |
        python3 -m venv .venv
        .venv/bin/pip install --quiet scipy numpy
        .venv/bin/python tests/scipy_ring_comparison.py --build-dir build --test-binary SequenceSolveTest
        .venv/bin/python tests/scipy_ring_comparison.py --build-dir build --test-binary MetalSequenceSolveTest

  # CUDA GPU tests on self-hosted NVIDIA runner
  cuda-gpu:
    runs-on: [nvidia-runner-1]
    container:
      image: nvidia/cuda:12.6.3-devel-ubuntu24.04
      options: --gpus all

    steps:
    - uses: actions/checkout@v4

    - name: Cache CMake FetchContent dependencies
      uses: actions/cache@v4
      with:
        path: build/_deps
        key: deps-${{ runner.os }}-${{ runner.arch }}-${{ hashFiles('CMakeLists.txt') }}
        restore-keys: |
          deps-${{ runner.os }}-${{ runner.arch }}-

    - name: Install dependencies
      run: |
        apt-get update
        apt-get install -y cmake build-essential libopenblas-dev python3-pip

    - name: System info
      run: |
        echo "=== GPU Info ==="
        nvidia-smi || echo "nvidia-smi not available"
        echo "=== CUDA ==="
        nvcc --version
        echo "=== CMake ==="
        cmake --version

    - name: Configure CMake
      run: |
        cmake -S . -B build \
          -DCMAKE_BUILD_TYPE=${{ env.BUILD_TYPE }} \
          -DSPRUX_USE_CUBLAS=ON \
          -DSPRUX_USE_METAL=OFF \
          -DSPRUX_USE_OPENCL=OFF \
          -DSPRUX_BUILD_TESTS=ON \
          -DSPRUX_BUILD_EXAMPLES=ON

    - name: Build
      run: cmake --build build --config ${{ env.BUILD_TYPE }} -j"$(nproc)"

    - name: Run all tests
      run: ctest --test-dir build --output-on-failure -j"$(nproc)"

    - name: Scipy cross-validation (CPU + CUDA)
      run: |
        apt-get install -y -qq python3-venv
        python3 -m venv .venv
        .venv/bin/pip install --quiet scipy numpy
        .venv/bin/python tests/scipy_ring_comparison.py --build-dir build --test-binary SequenceSolveTest
        .venv/bin/python tests/scipy_ring_comparison.py --build-dir build --test-binary CudaSequenceSolveTest