name: Unit tests

on:
  workflow_dispatch:
  schedule:
    # Every day at 02:15 AM UTC
    - cron: "15 2 * * *"
  push:
    branches: [testing-ci]

concurrency:
  group: ${{ github.workflow }}-${{ github.ref }}
  cancel-in-progress: true

jobs:

  build-cpu:
    strategy:
      matrix:
        os: [ubuntu-22.04, ubuntu-22.04-arm, windows-2025]
        include:
          - os: ubuntu-22.04
            arch: x86_64
          - os: ubuntu-22.04-arm
            arch: aarch64
          - os: windows-2025
            arch: x86_64
    runs-on: ${{ matrix.os }}
    steps:
      - uses: actions/checkout@v4

      - name: Setup MSVC
        if: startsWith(matrix.os, 'windows')
        uses: ilammy/msvc-dev-cmd@v1.13.0 # to use cl

      - name: Build C++
        run: bash .github/scripts/build-cpu.sh
        env:
          build_os: ${{ matrix.os }}
          build_arch: ${{ matrix.arch }}

      - name: Upload build artifact
        uses: actions/upload-artifact@v4
        with:
          name: lib_cpu_${{ matrix.os }}_${{ matrix.arch }}
          path: output/${{ matrix.os }}/${{ matrix.arch }}/*
          retention-days: 7

  build-cuda:
    strategy:
      matrix:
        cuda_version: ["11.8.0", "12.8.1"]
        os: [ubuntu-22.04, ubuntu-22.04-arm, windows-2025]
        include:
          - os: ubuntu-22.04
            arch: x86_64
          - os: ubuntu-22.04-arm
            arch: aarch64
          - os: windows-2025
            arch: x86_64
    runs-on: ${{ matrix.os }}

    steps:
      - uses: actions/checkout@v4

      - name: Install CUDA Toolkit
        uses: Jimver/cuda-toolkit@v0.2.23
        if: startsWith(matrix.os, 'windows')
        id: cuda-toolkit
        with:
          cuda: ${{ matrix.cuda_version }}
          method: "network"
          sub-packages: '["nvcc","cudart","cusparse","cublas","thrust","nvrtc_dev","cublas_dev","cusparse_dev"]'
          use-github-cache: false

      - name: Setup MSVC
        if: startsWith(matrix.os, 'windows')
        uses: ilammy/msvc-dev-cmd@v1.13.0 # to use cl

      # We're running on T4 only for now, so we only target sm75.
      - name: Build C++ / CUDA
        run: bash .github/scripts/build-cuda.sh
        env:
          build_os: ${{ matrix.os }}
          build_arch: ${{ matrix.arch }}
          cuda_version: ${{ matrix.cuda_version }}
          cuda_targets: "75"

      - name: Upload build artifact
        uses: actions/upload-artifact@v4
        with:
          name: lib_cuda_${{matrix.cuda_version}}_${{ matrix.os }}_${{ matrix.arch }}
          path: output/${{ matrix.os }}/${{ matrix.arch }}/*
          retention-days: 7

  cpu-tests:
    needs: build-cpu
    strategy:
      fail-fast: false
      matrix:
        os: [ubuntu-22.04, ubuntu-22.04-arm, windows-2025]
        torch_version: ["2.7.0"]
        include:
          - os: ubuntu-22.04
            arch: x86_64
          - os: ubuntu-22.04-arm
            arch: aarch64
          - os: windows-2025
            arch: x86_64
    runs-on: ${{ matrix.os }}
    env:
      BNB_TEST_DEVICE: cpu
    steps:
      - uses: actions/checkout@v4

      - name: Download build artifact
        uses: actions/download-artifact@v4
        with:
          name: lib_cpu_${{ matrix.os }}_${{ matrix.arch }}
          path: bitsandbytes/
          merge-multiple: true

      - name: Setup Python
        uses: actions/setup-python@v5
        with:
          python-version: 3.9

      - name: Install dependencies
        run: |
          pip install torch==${{ matrix.torch_version }} --index-url https://download.pytorch.org/whl/cpu
          pip install -e ".[test]"
          pip install pytest-cov

      - name: Show installed packages
        run: pip list

      - name: Run tests
        run: pytest

  cuda-tests:
    needs: build-cuda
    strategy:
      fail-fast: false
      matrix:
        os: [ubuntu-22.04, windows-2025]
        arch: [x86_64]
        cuda_version: ["11.8.0", "12.8.1"]
        include:
          - cuda_version: "11.8.0"
            torch_version: "2.4.1"
            pypi_index: "https://download.pytorch.org/whl/cu118"
          - cuda_version: "12.8.1"
            torch_version: "2.7.0"
            pypi_index: "https://download.pytorch.org/whl/cu128"
        exclude:
          # Our current T4 Windows runner has a driver too old (471.11)
          # and cannot support CUDA 12+. Skip for now.
          - os: windows-2025
            cuda_version: "12.8.1"
    runs-on:
      labels: ${{ contains(matrix.os, 'windows') && 'CUDA-Windows-x64' || 'CUDA-Linux-x64' }}
    env:
      BNB_TEST_DEVICE: cuda
    steps:
      - name: Show GPU Information
        run: nvidia-smi

      - uses: actions/checkout@v4

      - name: Download build artifact
        uses: actions/download-artifact@v4
        with:
          name: lib_cuda_${{ matrix.cuda_version }}_${{ matrix.os }}_${{ matrix.arch }}
          path: bitsandbytes/
          merge-multiple: true

      - name: Setup Python
        uses: actions/setup-python@v5
        with:
          python-version: 3.9

      - name: Install dependencies
        run: |
          pip install torch==${{ matrix.torch_version }} --index-url ${{ matrix.pypi_index }}
          pip install -e ".[test]"
          pip install pytest-cov

      - name: Show installed packages
        run: pip list

      - name: Run tests
        run: pytest