refactor(gpu): Specify launch bounds on kernels #816
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
| # Compile and test tfhe-cuda-backend on an AWS instance | |
| name: TFHE Cuda Backend - Full tests multi-GPU | |
| env: | |
| CARGO_TERM_COLOR: always | |
| ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }} | |
| RUSTFLAGS: "-C target-cpu=native" | |
| RUST_BACKTRACE: "full" | |
| RUST_MIN_STACK: "8388608" | |
| SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }} | |
| SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png | |
| SLACK_USERNAME: ${{ secrets.BOT_USERNAME }} | |
| SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }} | |
| IS_PULL_REQUEST: ${{ github.event_name == 'pull_request' }} | |
| on: | |
| # Allows you to run this workflow manually from the Actions tab as an alternative. | |
| workflow_dispatch: | |
| pull_request: | |
| types: [ labeled ] | |
| jobs: | |
| should-run: | |
| runs-on: ubuntu-latest | |
| permissions: | |
| pull-requests: write | |
| outputs: | |
| gpu_test: ${{ env.IS_PULL_REQUEST == 'false' || steps.changed-files.outputs.gpu_any_changed }} | |
| steps: | |
| - name: Checkout tfhe-rs | |
| uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332 | |
| with: | |
| fetch-depth: 0 | |
| - name: Check for file changes | |
| id: changed-files | |
| uses: tj-actions/changed-files@6b2903bdce6310cfbddd87c418f253cf29b2dec9 | |
| with: | |
| since_last_remote_commit: true | |
| files_yaml: | | |
| gpu: | |
| - tfhe/Cargo.toml | |
| - tfhe/build.rs | |
| - backends/tfhe-cuda-backend/** | |
| - tfhe/src/core_crypto/gpu/** | |
| - tfhe/src/integer/gpu/** | |
| - tfhe/shortint/parameters/** | |
| - tfhe/src/high_level_api/** | |
| - tfhe/src/c_api/** | |
| - 'tfhe/docs/**.md' | |
| - Makefile | |
| - '.github/workflows/aws_tfhe_multi_gpu**' | |
| - scripts/** | |
| - ci/** | |
| setup-instance: | |
| name: Setup instance (cuda-tests-multi-gpu) | |
| needs: should-run | |
| if: github.event_name != 'pull_request' || | |
| (github.event.action != 'labeled' && needs.should-run.outputs.gpu_test == 'true') || | |
| (github.event.action == 'labeled' && github.event.label.name == 'approved' && needs.should-run.outputs.gpu_test == 'true') | |
| runs-on: ubuntu-latest | |
| outputs: | |
| runner-name: ${{ steps.start-instance.outputs.label }} | |
| steps: | |
| - name: Start instance | |
| id: start-instance | |
| uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261 | |
| with: | |
| mode: start | |
| github-token: ${{ secrets.SLAB_ACTION_TOKEN }} | |
| slab-url: ${{ secrets.SLAB_BASE_URL }} | |
| job-secret: ${{ secrets.JOB_SECRET }} | |
| backend: aws | |
| profile: multi-gpu-test | |
| cuda-tests-linux: | |
| name: CUDA multi-GPU tests | |
| needs: [ should-run, setup-instance ] | |
| if: github.event_name != 'pull_request' || | |
| (github.event_name == 'pull_request' && needs.setup-instance.result != 'skipped') | |
| concurrency: | |
| group: ${{ github.workflow }}_${{ github.ref }} | |
| cancel-in-progress: ${{ github.ref != 'refs/heads/main' }} | |
| runs-on: ${{ needs.setup-instance.outputs.runner-name }} | |
| strategy: | |
| fail-fast: false | |
| # explicit include-based build matrix, of known valid options | |
| matrix: | |
| include: | |
| - os: ubuntu-22.04 | |
| cuda: "12.2" | |
| gcc: 9 | |
| env: | |
| CUDA_PATH: /usr/local/cuda-${{ matrix.cuda }} | |
| steps: | |
| - name: Checkout tfhe-rs | |
| uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332 | |
| - name: Set up home | |
| run: | | |
| echo "HOME=/home/ubuntu" >> "${GITHUB_ENV}" | |
| - name: Install latest stable | |
| uses: dtolnay/rust-toolchain@21dc36fb71dd22e3317045c0c31a3f4249868b17 | |
| with: | |
| toolchain: stable | |
| - name: Export CUDA variables | |
| if: ${{ !cancelled() }} | |
| run: | | |
| echo "CUDA_PATH=$CUDA_PATH" >> "${GITHUB_ENV}" | |
| echo "$CUDA_PATH/bin" >> "${GITHUB_PATH}" | |
| echo "LD_LIBRARY_PATH=$CUDA_PATH/lib:$LD_LIBRARY_PATH" >> "${GITHUB_ENV}" | |
| echo "CUDACXX=/usr/local/cuda-${{ matrix.cuda }}/bin/nvcc" >> "${GITHUB_ENV}" | |
| # Specify the correct host compilers | |
| - name: Export gcc and g++ variables | |
| if: ${{ !cancelled() }} | |
| run: | | |
| { | |
| echo "CC=/usr/bin/gcc-${{ matrix.gcc }}"; | |
| echo "CXX=/usr/bin/g++-${{ matrix.gcc }}"; | |
| echo "CUDAHOSTCXX=/usr/bin/g++-${{ matrix.gcc }}"; | |
| echo "HOME=/home/ubuntu"; | |
| } >> "${GITHUB_ENV}" | |
| # No need to test core_crypto and classic PBS in integer since it's already tested on single GPU. | |
| - name: Run multi-bit CUDA integer tests | |
| run: | | |
| make test_integer_multi_bit_gpu_ci | |
| - name: Run user docs tests | |
| run: | | |
| make test_user_doc_gpu | |
| - name: Test C API | |
| run: | | |
| make test_c_api_gpu | |
| - name: Run High Level API Tests | |
| run: | | |
| make test_high_level_api_gpu | |
| - name: Slack Notification | |
| if: ${{ always() }} | |
| continue-on-error: true | |
| uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907 | |
| env: | |
| SLACK_COLOR: ${{ job.status }} | |
| SLACK_MESSAGE: "CUDA AWS multi-GPU tests finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})" | |
| teardown-instance: | |
| name: Teardown instance (cuda-tests-multi-gpu) | |
| if: ${{ always() && needs.setup-instance.result != 'skipped' }} | |
| needs: [ setup-instance, cuda-tests-linux ] | |
| runs-on: ubuntu-latest | |
| steps: | |
| - name: Stop instance | |
| id: stop-instance | |
| uses: zama-ai/slab-github-runner@447a2d0fd2d1a9d647aa0d0723a6e9255372f261 | |
| with: | |
| mode: stop | |
| github-token: ${{ secrets.SLAB_ACTION_TOKEN }} | |
| slab-url: ${{ secrets.SLAB_BASE_URL }} | |
| job-secret: ${{ secrets.JOB_SECRET }} | |
| label: ${{ needs.setup-instance.outputs.runner-name }} | |
| - name: Slack Notification | |
| if: ${{ failure() }} | |
| continue-on-error: true | |
| uses: rtCamp/action-slack-notify@4e5fb42d249be6a45a298f3c9543b111b02f7907 | |
| env: | |
| SLACK_COLOR: ${{ job.status }} | |
| SLACK_MESSAGE: "Instance teardown (cuda-tests-multi-gpu) finished with status: ${{ job.status }}. (${{ env.ACTION_RUN_URL }})" |