release-python.yml

name: release-python
on:
  push:
    tags:
      - '*'
  workflow_dispatch:

env:
  DEVELOPER_DIR: /Applications/Xcode_16.4.0.app/Contents/Developer
  MAC_DEPLOYMENT_TARGET: '11.0'
  ENABLE_BITCODE: OFF
  ENABLE_ARC: OFF
  ENABLE_VISIBILITY: OFF
  CIBW_SKIP: "cp3??t-*"

jobs:
  build_sdist:
    name: Build SDist
    runs-on: ubuntu-latest
    steps:
    - uses: actions/checkout@v7
      with:
        submodules: true

    - uses: actions/setup-python@v7
      with:
        python-version: '3.x'

    - name: Install deps
      run: python -m pip install twine build

    - name: Build SDist
      run: python -m build -s

    - name: Check metadata
      run: twine check dist/*

    - uses: actions/upload-artifact@v7
      with:
        name: sdist
        path: dist/*.tar.gz

  build_wheels:
    name: ${{ matrix.arch }} ${{ matrix.build_id }} on ${{ matrix.os }}
    runs-on: ${{ matrix.os }}
    strategy:
      fail-fast: false
      matrix:
        include:
          - { os: ubuntu-24.04,     arch: x86_64,     build: 'cp*-manylinux*', build_id: cp-manylinux }
          - { os: ubuntu-24.04,     arch: x86_64,     build: 'cp*-musllinux*', build_id: cp-musllinux }
          - { os: ubuntu-24.04,     arch: x86_64,     build: 'pp*',            build_id: pp           }
          - { os: ubuntu-24.04,     arch: i686,       build: 'cp*-manylinux*', build_id: cp-manylinux }
          - { os: ubuntu-24.04,     arch: i686,       build: 'cp*-musllinux*', build_id: cp-musllinux }
          - { os: ubuntu-24.04,     arch: i686,       build: 'pp*',            build_id: pp           }
          - { os: windows-2025,     arch: x86,        build: 'cp*',            build_id: cp           }
          - { os: windows-2025,     arch: AMD64,      build: 'cp*',            build_id: cp           }
          - { os: windows-2025,     arch: AMD64,      build: 'pp*',            build_id: pp           }
          - { os: windows-11-arm,   arch: ARM64,      build: 'cp*',            build_id: cp           }
          - { os: macos-15-intel,   arch: x86_64,     build: 'cp*',            build_id: cp           }
          - { os: macos-15,         arch: arm64,      build: 'cp*',            build_id: cp           }
          - { os: ubuntu-24.04-arm, arch: armv7l,     build: 'cp*-manylinux*', build_id: cp-manylinux }
          - { os: ubuntu-24.04-arm, arch: armv7l,     build: 'cp*-musllinux*', build_id: cp-musllinux }
          - { os: ubuntu-24.04-arm, arch: aarch64,    build: 'cp*-manylinux*', build_id: cp-manylinux }
          - { os: ubuntu-24.04-arm, arch: aarch64,    build: 'cp*-musllinux*', build_id: cp-musllinux }
          - { os: ubuntu-24.04-arm, arch: aarch64,    build: 'pp*',            build_id: pp           }

    env:
      OPENMP_VERSION: '18.1.2'
      OPENMP_CMAKE_OPTIONS: |
        -DCMAKE_TOOLCHAIN_FILE=../../toolchains/ios.toolchain.cmake \
        -DDEPLOYMENT_TARGET=$MAC_DEPLOYMENT_TARGET \
        -DENABLE_BITCODE=$ENABLE_BITCODE \
        -DENABLE_ARC=$ENABLE_ARC \
        -DENABLE_VISIBILITY=$ENABLE_VISIBILITY \
        -DCMAKE_INSTALL_PREFIX=install \
        -DCMAKE_BUILD_TYPE=Release \
        -DLIBOMP_ENABLE_SHARED=OFF \
        -DLIBOMP_OMPT_SUPPORT=OFF \
        -DLIBOMP_USE_HWLOC=OFF \

    steps:
    - uses: actions/checkout@v7
      with:
        submodules: true

    # build wheels for ubuntu
    - name: Build wheels for ubuntu
      if: matrix.os == 'ubuntu-24.04'
      uses: pypa/cibuildwheel@v4.3.5
      env:
        CIBW_ARCHS_LINUX: ${{ matrix.arch }}
        CIBW_BUILD: ${{ matrix.build }}
        CIBW_ENABLE: pypy
        CIBW_BUILD_VERBOSITY: 1
        CIBW_ENVIRONMENT: CMAKE_BUILD_PARALLEL_LEVEL=4
      with:
        output-dir: wheelhouse

    # build wheels for ubuntu armv7l
    - name: Build wheels for ubuntu armv7l
      if: matrix.os == 'ubuntu-24.04-arm' && (matrix.arch == 'armv7l')
      uses: pypa/cibuildwheel@v4.3.5
      env:
        CIBW_ARCHS_LINUX: ${{ matrix.arch }}
        CIBW_BUILD: ${{ matrix.build }}
        CIBW_ENABLE: pypy
        CIBW_BUILD_VERBOSITY: 1
        CIBW_ENVIRONMENT: CMAKE_BUILD_PARALLEL_LEVEL=4
          CFLAGS="-mfpu=neon" CXXFLAGS="-mfpu=neon"
      with:
        output-dir: wheelhouse

    # build wheels for ubuntu aarch64
    - name: Build wheels for ubuntu aarch64
      if: matrix.os == 'ubuntu-24.04-arm' && (matrix.arch == 'aarch64')
      uses: pypa/cibuildwheel@v4.3.5
      env:
        CIBW_ARCHS_LINUX: ${{ matrix.arch }}
        CIBW_BUILD: ${{ matrix.build }}
        CIBW_ENABLE: pypy
        CIBW_BUILD_VERBOSITY: 1
        CIBW_ENVIRONMENT: CMAKE_BUILD_PARALLEL_LEVEL=4
      with:
        output-dir: wheelhouse

    # build wheels for windows
    - name: Build wheels for windows
      if: matrix.os == 'windows-2025' && (matrix.arch == 'AMD64' || matrix.arch == 'x86')
      uses: pypa/cibuildwheel@v4.3.5
      env:
        CIBW_ARCHS_WINDOWS: ${{ matrix.arch }}
        CIBW_BUILD: ${{ matrix.build }}
        CIBW_ENABLE: pypy
        CIBW_BUILD_VERBOSITY: 1
        CIBW_ENVIRONMENT_WINDOWS: CMAKE_BUILD_PARALLEL_LEVEL=4
        CIBW_BEFORE_BUILD: pip install delvewheel
        CIBW_REPAIR_WHEEL_COMMAND: delvewheel repair -w {dest_dir} {wheel}
      with:
        output-dir: wheelhouse

    - name: Build wheels for windows ARM64
      if: matrix.os == 'windows-11-arm' && matrix.arch == 'ARM64'
      uses: pypa/cibuildwheel@v4.3.5
      env:
        CIBW_ARCHS_WINDOWS: ${{ matrix.arch }}
        CIBW_BUILD: ${{ matrix.build }}
        CIBW_BUILD_VERBOSITY: 1
        CIBW_ENVIRONMENT_WINDOWS: CMAKE_BUILD_PARALLEL_LEVEL=4
        CIBW_BEFORE_BUILD: pip install delvewheel
        CIBW_REPAIR_WHEEL_COMMAND: delvewheel repair -w {dest_dir} {wheel} --no-dll "msvcp140.dll;vcomp140.dll"
      with:
        output-dir: wheelhouse

    # build wheels for macos
    - name: cache-openmp for macos
      if: matrix.os == 'macos-15-intel' || matrix.os == 'macos-15'
      id: cache-openmp
      uses: actions/cache@v6
      with:
        path: openmp-install
        key: openmp-macos-install-20251004

    - name: openmp for macos
      if: (matrix.os == 'macos-15-intel' || matrix.os == 'macos-15') && steps.cache-openmp.outputs.cache-hit != 'true'
      run: |
        wget https://github.com/llvm/llvm-project/releases/download/llvmorg-${{ env.OPENMP_VERSION }}/cmake-${{ env.OPENMP_VERSION }}.src.tar.xz
        tar -xf cmake-${{ env.OPENMP_VERSION }}.src.tar.xz
        wget https://github.com/llvm/llvm-project/releases/download/llvmorg-${{ env.OPENMP_VERSION }}/openmp-${{ env.OPENMP_VERSION }}.src.tar.xz
        tar -xf openmp-${{ env.OPENMP_VERSION }}.src.tar.xz
        mv cmake-${{ env.OPENMP_VERSION }}.src/Modules/* openmp-${{ env.OPENMP_VERSION }}.src/cmake/
        cd openmp-${{ env.OPENMP_VERSION }}.src
        wget https://github.com/nihui/llvm-project/commit/ef8c35bcf5d9cfdb0764ffde6a63c04ec715bc37.patch
        patch -p2 -i ef8c35bcf5d9cfdb0764ffde6a63c04ec715bc37.patch
        wget https://github.com/nihui/llvm-project/commit/5c12711f9a21f41bea70566bf15a4026804d6b20.patch
        patch -p2 -i 5c12711f9a21f41bea70566bf15a4026804d6b20.patch

    - name: openmp-build-x86_64 for macos
      if: (matrix.os == 'macos-15-intel' || matrix.os == 'macos-15') && steps.cache-openmp.outputs.cache-hit != 'true'
      run: |
        cd openmp-${{ env.OPENMP_VERSION }}.src
        mkdir -p build-x86_64 && cd build-x86_64
        cmake ${{ env.OPENMP_CMAKE_OPTIONS }} -DPLATFORM=MAC -DARCHS="x86_64" ..
        cmake --build . -j 4
        cmake --build . --target install

    - name: openmp-build-arm64 for macos
      if: (matrix.os == 'macos-15-intel' || matrix.os == 'macos-15') && steps.cache-openmp.outputs.cache-hit != 'true'
      run: |
        cd openmp-${{ env.OPENMP_VERSION }}.src
        mkdir -p build-arm64 && cd build-arm64
        cmake ${{ env.OPENMP_CMAKE_OPTIONS }} -DPLATFORM=MAC_ARM64 -DARCHS="arm64" ..
        cmake --build . -j 4
        cmake --build . --target install

    - name: openmp-merge-fat-library for macos
      if: (matrix.os == 'macos-15-intel' || matrix.os == 'macos-15') && steps.cache-openmp.outputs.cache-hit != 'true'
      run: |
        mkdir -p $GITHUB_WORKSPACE/openmp-install
        cp -a openmp-${{ env.OPENMP_VERSION }}.src/build-x86_64/install/include $GITHUB_WORKSPACE/openmp-install
        mkdir -p $GITHUB_WORKSPACE/openmp-install/lib
        lipo -create \
            openmp-${{ env.OPENMP_VERSION }}.src/build-x86_64/install/lib/libomp.a \
            openmp-${{ env.OPENMP_VERSION }}.src/build-arm64/install/lib/libomp.a \
            -o $GITHUB_WORKSPACE/openmp-install/lib/libomp.a

    - name: install-openmp for macos
      if: matrix.os == 'macos-15-intel' || matrix.os == 'macos-15'
      run: |
        sudo cp $GITHUB_WORKSPACE/openmp-install/include/* $DEVELOPER_DIR/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/include
        sudo cp $GITHUB_WORKSPACE/openmp-install/lib/libomp.a $DEVELOPER_DIR/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/lib

    - name: vulkansdk for macos
      if: matrix.os == 'macos-15-intel' || matrix.os == 'macos-15'
      run: |
        wget -q https://sdk.lunarg.com/sdk/download/1.4.335.1/mac/vulkansdk-macos-1.4.335.1.zip?Human=true -O vulkansdk-macos-1.4.335.1.zip
        unzip vulkansdk-macos-1.4.335.1.zip
        sudo vulkansdk-macOS-1.4.335.1.app/Contents/MacOS/vulkansdk-macOS-1.4.335.1 --root $GITHUB_WORKSPACE/vulkansdk-macos-1.4.335.1 --accept-licenses --default-answer --confirm-command install

    - name: Build wheels for macos x86_64
      if: matrix.os == 'macos-15-intel' && matrix.arch == 'x86_64'
      uses: pypa/cibuildwheel@v4.3.5
      env:
        CIBW_ARCHS_MACOS: ${{ matrix.arch }}
        CIBW_BUILD: ${{ matrix.build }}
        CIBW_ENABLE: pypy
        CIBW_BUILD_VERBOSITY: 1
        CIBW_ENVIRONMENT: CMAKE_BUILD_PARALLEL_LEVEL=4
          CMAKE_TOOLCHAIN_FILE=$GITHUB_WORKSPACE/toolchains/ios.toolchain.cmake PLATFORM=MAC ARCHS="x86_64"
          DEPLOYMENT_TARGET=$MAC_DEPLOYMENT_TARGET ENABLE_BITCODE=OFF ENABLE_ARC=OFF ENABLE_VISIBILITY=OFF
          OpenMP_C_FLAGS="-Xclang -fopenmp" OpenMP_CXX_FLAGS="-Xclang -fopenmp"
          OpenMP_C_LIB_NAMES="libomp" OpenMP_CXX_LIB_NAMES="libomp"
          OpenMP_libomp_LIBRARY="libomp.a"
          Vulkan_LIBRARY=$GITHUB_WORKSPACE/vulkansdk-macos-1.4.335.1/macOS/lib/libMoltenVK.dylib
          MACOSX_DEPLOYMENT_TARGET=$MAC_DEPLOYMENT_TARGET
      with:
        output-dir: wheelhouse

    - name: Build wheels for macos arm64
      if: matrix.os == 'macos-15' && matrix.arch == 'arm64'
      uses: pypa/cibuildwheel@v4.3.5
      env:
        CIBW_ARCHS_MACOS: ${{ matrix.arch }}
        CIBW_BUILD: ${{ matrix.build }}
        CIBW_ENABLE: pypy
        CIBW_BUILD_VERBOSITY: 1
        CIBW_ENVIRONMENT: CMAKE_BUILD_PARALLEL_LEVEL=4
          CMAKE_TOOLCHAIN_FILE=$GITHUB_WORKSPACE/toolchains/ios.toolchain.cmake PLATFORM=MAC_ARM64 ARCHS="arm64"
          DEPLOYMENT_TARGET=$MAC_DEPLOYMENT_TARGET ENABLE_BITCODE=OFF ENABLE_ARC=OFF ENABLE_VISIBILITY=OFF
          OpenMP_C_FLAGS="-Xclang -fopenmp" OpenMP_CXX_FLAGS="-Xclang -fopenmp"
          OpenMP_C_LIB_NAMES="libomp" OpenMP_CXX_LIB_NAMES="libomp"
          OpenMP_libomp_LIBRARY="libomp.a"
          Vulkan_LIBRARY=$GITHUB_WORKSPACE/vulkansdk-macos-1.4.335.1/macOS/lib/libMoltenVK.dylib
          MACOSX_DEPLOYMENT_TARGET=$MAC_DEPLOYMENT_TARGET
      with:
        output-dir: wheelhouse

    - name: Show files
      run: ls -lh wheelhouse
      shell: bash

    - name: Verify clean directory
      run: git diff --exit-code
      shell: bash

    - name: Upload wheels
      uses: actions/upload-artifact@v7
      with:
        name: wheels-${{ matrix.os }}-${{ matrix.arch }}-${{ matrix.build_id }}
        path: wheelhouse/*.whl

  build_wheels_qemu_cp:
    name: ${{ matrix.arch }} ${{ matrix.build_cp }} ${{ matrix.build_sub }}
    runs-on: ubuntu-24.04

    strategy:
      fail-fast: false
      matrix:
        arch: [riscv64]
        build_cp: [cp39, cp310, cp311, cp312, cp313, cp314, cp315]
        build_sub: [manylinux, musllinux]

    steps:
    - uses: actions/checkout@v7
      with:
        submodules: true

    - name: Set up QEMU
      uses: docker/setup-qemu-action@v4
      with:
        platforms: all

    - name: Build wheels with qemu
      uses: pypa/cibuildwheel@v4.3.5
      env:
        CIBW_ARCHS_LINUX: ${{ matrix.arch }}
        CIBW_BUILD: ${{ matrix.build_cp }}-${{ matrix.build_sub }}*
        CIBW_BUILD_VERBOSITY: 1
        CIBW_ENVIRONMENT: CMAKE_BUILD_PARALLEL_LEVEL=4 EXTRA_CMAKE_ARGS="-DNCNN_XTHEADVECTOR=OFF"
      with:
        output-dir: wheelhouse

    - name: Show files
      run: ls -lh wheelhouse
      shell: bash

    - name: Verify clean directory
      run: git diff --exit-code
      shell: bash

    - name: Upload wheels
      uses: actions/upload-artifact@v7
      with:
        name: wheels_qemu_cp-${{ matrix.arch }}-${{ matrix.build_cp }}-${{ matrix.build_sub }}
        path: wheelhouse/*.whl

  build_wheels_android:
    permissions:
      contents: read
    name: ${{ matrix.arch }} ${{ matrix.build_cp }} on Android
    runs-on: ubuntu-24.04
    timeout-minutes: 90
    strategy:
      fail-fast: false
      matrix:
        build_cp: [cp313, cp314, cp315]
        arch: [arm64_v8a, x86_64]
    steps:
    - uses: actions/checkout@v7
      with:
        submodules: true
    - uses: actions/setup-python@v7
      with:
        python-version: '3.x'
    # ubuntu's android sdk and cibuildwheel provision the matching ndk and
    # gradle-managed emulator. keep cibuildwheel's generated toolchain file
    - name: Build Android wheel and test native architecture
      uses: pypa/cibuildwheel@v4.3.5
      with:
        package-dir: .
        output-dir: wheelhouse
      env:
        CIBW_PLATFORM: android
        CIBW_ARCHS_ANDROID: ${{ matrix.arch }}
        CIBW_BUILD: ${{ matrix.build_cp }}-android_*
        CIBW_BUILD_FRONTEND: build
        CIBW_BUILD_VERBOSITY: 1
        CIBW_ENVIRONMENT_ANDROID: >-
          CMAKE_BUILD_PARALLEL_LEVEL=4
          EXTRA_CMAKE_ARGS="-DNCNN_VULKAN=OFF -DNCNN_OPENMP=OFF"
        # test the native binding without installing runtime dependencies
        # this only changes the disposable test venv, not wheel metadata
        CIBW_BEFORE_TEST_ANDROID: python -m pip config --site set install.no-deps true
        CIBW_TEST_SOURCES: python/tests/mobile_smoke.py
        CIBW_TEST_COMMAND: python -m unittest discover -s python/tests -p mobile_smoke.py -v
        # ubuntu x86_64 can emulator-test x86_64 only. arm64 remains build-only
        CIBW_TEST_SKIP: '*-android_arm64_v8a'
    - uses: actions/upload-artifact@v7
      with:
        name: wheels_android-${{ matrix.build_cp }}-${{ matrix.arch }}
        path: wheelhouse/*.whl
        if-no-files-found: error

  build_wheels_ios:
    permissions:
      contents: read
    name: ${{ matrix.arch }} ${{ matrix.build_cp }} on iOS
    runs-on: ${{ matrix.os }}
    timeout-minutes: 90
    strategy:
      fail-fast: false
      matrix:
        build_cp: [cp313, cp314, cp315]
        arch: [arm64_iphoneos, arm64_iphonesimulator, x86_64_iphonesimulator]
        include:
          - { os: macos-15, arch: arm64_iphoneos }
          - { os: macos-15, arch: arm64_iphonesimulator }
          - { os: macos-15-intel, arch: x86_64_iphonesimulator }
    steps:
    - uses: actions/checkout@v7
      with:
        submodules: true
    - uses: actions/setup-python@v7
      with:
        python-version: '3.x'
    - name: Install host cross-build tools
      run: |
        python -m pip install 'cmake>=4,<5' ninja
        cmake --version
        xcodebuild -version
        xcodebuild -showsdks
        xcrun simctl list devices available
    - name: Build iOS wheel and test native simulator
      uses: pypa/cibuildwheel@v4.3.5
      with:
        package-dir: .
        output-dir: wheelhouse
      env:
        CIBW_PLATFORM: ios
        CIBW_ARCHS_IOS: ${{ matrix.arch }}
        # cross sysconfig may report ios13 despite the deployment target override
        CIBW_CONFIG_SETTINGS_IOS: --build-option=--plat-name=ios_14_0_${{ matrix.arch }}
        CIBW_BUILD: ${{ matrix.build_cp }}-ios_*
        CIBW_BUILD_FRONTEND: build
        CIBW_BUILD_VERBOSITY: 1
        CIBW_XBUILD_TOOLS_IOS: cmake ninja
        CIBW_ENVIRONMENT_IOS: >-
          CMAKE_BUILD_PARALLEL_LEVEL=4
          CMAKE_GENERATOR=Ninja
          IPHONEOS_DEPLOYMENT_TARGET=14.0
          EXTRA_CMAKE_ARGS="-DNCNN_VULKAN=OFF -DNCNN_OPENMP=OFF"
        # test the native binding without installing runtime dependencies
        # this only changes the disposable testbed, not wheel metadata
        CIBW_TEST_ENVIRONMENT_IOS: PIP_NO_DEPS=1
        CIBW_TEST_SOURCES: python/tests/mobile_smoke.py
        CIBW_TEST_COMMAND: python -m unittest discover -s python/tests -p mobile_smoke.py -v
        # cibuildwheel can test only native simulators, never physical devices
        CIBW_TEST_SKIP: '*-ios_arm64_iphoneos'
    - name: Inspect wheel target and linked libraries
      env:
        EXPECTED_ARCH: ${{ matrix.arch }}
        EXPECTED_PYTHON: ${{ matrix.build_cp }}
      run: |
        python - <<'PY'
        import os, pathlib, re, subprocess, tempfile, zipfile

        arch, python = os.environ["EXPECTED_ARCH"], os.environ["EXPECTED_PYTHON"]
        cpu, sdk = arch.rsplit("_", 1)
        platform = "IOS" if sdk == "iphoneos" else "IOSSIMULATOR"
        wheels = list(pathlib.Path("wheelhouse").glob("*.whl"))
        assert len(wheels) == 1, wheels
        assert wheels[0].name.endswith(f"-{python}-{python}-ios_14_0_{arch}.whl"), wheels[0]
        with tempfile.TemporaryDirectory() as temp, zipfile.ZipFile(wheels[0]) as wheel:
            modules = [name for name in wheel.namelist() if name.endswith(".so")]
            expected = f"ncnn/ncnn.cpython-{python[2:]}-{sdk}.so"
            assert modules == [expected], modules
            module = wheel.extract(expected, temp)
            cpus = subprocess.check_output(["lipo", "-archs", module], text=True).split()
            assert cpus == [cpu], cpus
            build = subprocess.check_output(["xcrun", "vtool", "-show-build", module], text=True)
            print(build)
            assert re.search(r"\bplatform\s+" + platform + r"\b", build), build
            minimum = re.search(r"\bminos\s+(\d+(?:\.\d+){0,2})\b", build)
            assert minimum, "missing Mach-O deployment target"
            version = tuple(map(int, minimum.group(1).split(".")))
            assert (version + (0, 0, 0))[:3] <= (14, 0, 0), minimum.group(1)
            links = subprocess.check_output(["otool", "-L", module], text=True)
            print(links)
            assert not any(path in links for path in ("/opt/homebrew/", "/usr/local/", "MacOSX.sdk")), links
        PY
    - uses: actions/upload-artifact@v7
      with:
        name: wheels_ios-${{ matrix.build_cp }}-${{ matrix.arch }}
        path: wheelhouse/*.whl
        if-no-files-found: error

  upload_all:
    permissions:
      contents: none
    name: Upload
    needs: [build_wheels, build_wheels_qemu_cp, build_wheels_android, build_wheels_ios, build_sdist]
    runs-on: ubuntu-latest

    steps:
    - uses: actions/download-artifact@v8
      with:
        path: dist
        merge-multiple: true

    - uses: pypa/gh-action-pypi-publish@release/v1
      with:
        user: __token__
        password: ${{ secrets.PYPI_API_TOKEN }}