LLM GEMM benchmarks #382
Workflow file for this run
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: "Test" | |
| on: | |
| push: | |
| branches: [ "sycl-develop" ] | |
| pull_request: | |
| branches: [ "sycl-develop" ] | |
| workflow_dispatch: | |
| inputs: | |
| DPCPP_VERSION: | |
| description: "DPCPP version to use" | |
| type: string | |
| permissions: {} | |
| jobs: | |
| run-tests: | |
| name: Run tests | |
| runs-on: cp-nvidia-gpu | |
| timeout-minutes: 30 | |
| steps: | |
| - name: Checkout repository | |
| uses: actions/checkout@a5ac7e51b41094c92402da3b24376905380afc29 # v4.1.6 | |
| - name: Install dependencies | |
| shell: bash | |
| run: | | |
| nvidia-smi | |
| mkdir ~/dpcpp | |
| pushd ~/dpcpp | |
| echo "Will use DPCPP ${{ inputs.DPCPP_VERSION }}." | |
| if [[ "${{ inputs.DPCPP_VERSION }}" != "" ]]; then \ | |
| echo "Downloading DPCPP from https://github.com/intel/llvm/releases/download/${{ inputs.DPCPP_VERSION }}/sycl_linux.tar.gz"; \ | |
| wget -q https://github.com/intel/llvm/releases/download/${{ inputs.DPCPP_VERSION }}/sycl_linux.tar.gz; \ | |
| else | |
| latest=$(curl -sS https://api.github.com/repos/intel/llvm/releases | jq -r '[.[].tag_name|select(match("nightly-[0-9]{4}-[0-9]{2}-[0-9]{2}"))][0]') && \ | |
| echo "Downloading DPCPP from https://github.com/intel/llvm/releases/download/${latest}/sycl_linux.tar.gz"; \ | |
| wget -q https://github.com/intel/llvm/releases/download/${latest}/sycl_linux.tar.gz; \ | |
| fi | |
| tar -xf sycl_linux.tar.gz | |
| rm sycl_linux.tar.gz | |
| popd | |
| # For a specific DPC++ nightly build define the repository variable DPCPP_VERSION | |
| # for example using the tag: 'nightly-2024-04-22' | |
| - name: Build | |
| shell: bash | |
| run: | | |
| export PATH=~/dpcpp/bin/:$PATH | |
| export C_INCLUDE_PATH=~/dpcpp/include/:$C_INCLUDE_PATH | |
| export CPLUS_INCLUDE_PATH=~/dpcpp/include/:$CPLUS_INCLUDE_PATH | |
| export LD_LIBRARY_PATH=~/dpcpp/lib/:$LD_LIBRARY_PATH | |
| export CC=clang | |
| export CXX=clang++ | |
| clang++ --version | |
| cmake -G Ninja \ | |
| -DCMAKE_CUDA_HOST_COMPILER=clang++ \ | |
| -DCUTLASS_ENABLE_SYCL=ON \ | |
| -DDPCPP_SYCL_TARGET=nvptx64-nvidia-cuda \ | |
| -DDPCPP_SYCL_ARCH=sm_80 | |
| cmake --build . | |
| - name: Unit test | |
| shell: bash | |
| run: | | |
| export LD_LIBRARY_PATH=~/dpcpp/lib/:$LD_LIBRARY_PATH | |
| cmake --build . --target test_unit -j 24 | |
| - name: Examples | |
| shell: bash | |
| run: | | |
| export LD_LIBRARY_PATH=~/dpcpp/lib/:$LD_LIBRARY_PATH | |
| echo Run sgemm_1 | |
| ./examples/cute/tutorial/sgemm_1 | |
| echo Run sgemm_2 | |
| ./examples/cute/tutorial/sgemm_2 | |
| echo Run sgemm_sm70 | |
| ./examples/cute/tutorial/sgemm_sm70 | |
| echo Run sgemm_sm80 | |
| ./examples/cute/tutorial/sgemm_sm80 | |
| echo Run tiled_copy | |
| ./examples/cute/tutorial/tiled_copy |