windows-trt.yml

name: Build (Windows-TRT)

on:
  push:
    paths:
      - 'vstrt/**'
      - '.github/workflows/windows-trt.yml'
  workflow_call:
    inputs:
      tag:
        description: 'which tag to upload to'
        required: true
        type: string
    secrets:
      REPO_TOKEN:
        required: true
  workflow_dispatch:
    inputs:
      tag:
        description: 'which tag to upload to'
        default: ''

jobs:
  build-windows:
    runs-on: windows-2022

    defaults:
      run:
        shell: cmd
        working-directory: vstrt

    steps:
    - name: Checkout repo
      uses: actions/checkout@v4
      with:
        fetch-depth: 0

    - name: Setup MSVC
      uses: ilammy/msvc-dev-cmd@v1

    - name: Setup Ninja
      run: pip install ninja

    - name: Cache CUDA
      id: cache-cuda
      uses: actions/cache@v4
      with:
        path: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA
        key: ${{ runner.os }}-vstrt-cuda-12.6.2
        save-always: true

    - name: Setup CUDA
      if: steps.cache-cuda.outputs.cache-hit != 'true'
      run: |
        curl -s -o cuda_installer.exe -L https://developer.download.nvidia.com/compute/cuda/12.6.2/network_installers/cuda_12.6.2_windows_network.exe
        cuda_installer.exe -s nvcc_12.6 cudart_12.6 cuda_profiler_api_12.6

    - name: Download TensorRT
      run: |
        curl -L -o trt.zip https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.6.0/zip/TensorRT-10.6.0.26.Windows.win10.cuda-12.6.zip
        unzip trt.zip
        mv TensorRT-*/ tensorrt/

    - name: Download VapourSynth headers
      run: |
        curl -s -o vs.zip -L https://github.com/vapoursynth/vapoursynth/archive/refs/tags/R54.zip
        unzip -q vs.zip
        mv vapoursynth-*/ vapoursynth/

    - name: Configure
      run: cmake -S . -B build -G Ninja -LA
        -D CMAKE_BUILD_TYPE=Release
        -D CMAKE_CXX_FLAGS="/EHsc /Wall /wd4100 /wd4625 /wd4626 /wd4710 /wd4711 /wd4820 /wd4996 /wd5026 /wd5027"
        -D CMAKE_MSVC_RUNTIME_LIBRARY=MultiThreaded
        -D CUDAToolkit_ROOT="C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.6"
        -D VAPOURSYNTH_INCLUDE_DIRECTORY="%cd%\vapoursynth\include"
        -D TENSORRT_HOME="%cd%\tensorrt"
        -D USE_NVINFER_PLUGIN=ON
        -D TENSORRT_LIBRARY_SUFFIX="_10"

    - name: Build
      run: cmake --build build --config Release --verbose

    - name: Install
      run: cmake --install build --prefix install

    - name: Checkout TensorRT OSS
      uses: actions/checkout@v4
      with:
        repository: NVIDIA/TensorRT
        ref: v10.1.0
        fetch-depth: 1
        path: tensorrt-oss

    - name: Override trtexec CMake file
      run: |
        cp -f -r -v tensorrt/samples ../tensorrt-oss
        cp -f -r -v tensorrt/include ../tensorrt-oss

        mv trtexec/CMakeLists.txt ../tensorrt-oss/samples/trtexec
        mv trtexec/*.cpp ../tensorrt-oss/samples/trtexec
        mv longpath.manifest ../tensorrt-oss/samples/trtexec

    - name: Apply patch
      run: |
        mv trtexec/trtexec.patch ../tensorrt-oss
        cd ../tensorrt-oss

        copy samples\utils\fileLock.cpp samples\utils\fileLock-utf16le.cpp
        powershell "Get-Content samples\utils\fileLock-utf16le.cpp | Out-File samples\utils\fileLock.cpp -Encoding ascii"
        git apply trtexec.patch --verbose
        copy samples\utils\fileLock.cpp samples\utils\fileLock-utf8.cpp
        powershell "Get-Content samples\utils\fileLock-utf8.cpp | Out-File samples\utils\fileLock.cpp"

    - name: Configure trtexec
      run: cmake -S ../tensorrt-oss/samples/trtexec -B build_trtexec -G Ninja
        -D CMAKE_BUILD_TYPE=Release
        -D CMAKE_MSVC_RUNTIME_LIBRARY=MultiThreaded
        -D CUDAToolkit_ROOT="C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.6"
        -D CMAKE_UNITY_BUILD=ON -D CMAKE_UNITY_BUILD_BATCH_SIZE=0

    - name: Build trtexec
      run: cmake --build build_trtexec --verbose

    - name: Install trtexec
      run: cmake --install build_trtexec --prefix trtexec

    - name: Prepare for upload
      run: |
        mkdir artifact
        copy install\bin\vstrt.dll artifact\
        mkdir artifact\vsmlrt-cuda
        copy trtexec\bin\trtexec.exe artifact\vsmlrt-cuda

    - name: Describe
      run: git describe --tags --long

    - name: Dump dependencies
      run: dumpbin /dependents artifact\vstrt.dll

    - name: Upload
      uses: actions/upload-artifact@v4
      with:
        name: VSTRT-Windows-x64
        path: vstrt/artifact

    - name: Compress artifact for release
      if: github.event_name == 'workflow_dispatch' && github.event.inputs.tag != ''
      run: |
        cd artifact
        7z a -t7z -mx=7 ../../VSTRT-Windows-x64.${{ github.event.inputs.tag }}.7z .

    - name: Release
      uses: softprops/action-gh-release@v2
      if: github.event_name == 'workflow_dispatch' && github.event.inputs.tag != ''
      with:
        tag_name: ${{ inputs.tag }}
        files: VSTRT-Windows-x64.${{ github.event.inputs.tag }}.7z
        fail_on_unmatched_files: true
        generate_release_notes: false
        prerelease: true