sel4bench

seL4 Benchmark Builds

This action builds sel4bench test images. The test runs can be restricted to specific arch, mode, and march settings.

It does so by doing a build from the default sel4bench-manifest for multiple configurations, advancing the sel4bench repository to the branch of the pull request the test is called on, and building an image for the selected platform.

Content

The entry point is the script steps.sh.

Build and platform configurations are in the respective yaml and python files in this repository. See these files for config documentation.

The main test driver is build.py in this directory.

Arguments

To add or modify build configurations, edit builds.yml in this directory. To filter the build variants defined there for a specific run, use one or more of the following:

• arch: comma separated list of architecture to filter on, e.g arm, riscv.
• march: comma separated list of march flags, e.g. armv7a, nehalem
• mode: one of {32, 64}
• xml: sel4bench xml manifest to test
• iterations: number of benchmark iterations to run, overrides the ITERATIONS setting in builds.yml.

Example

Put this into a .github/workflows/ yaml file, e.g. sel4bench.yml:

name: sel4bench

on: [pull_request]

jobs:
  build:
    name: Build
    runs-on: ubuntu-latest
    steps:
    strategy:
          matrix:
            march: [armv7a, armv8a, nehalem, rv32imac, rv64imac]
    steps:
    - uses: seL4/ci-actions/sel4bench@master
      with:
        march: ${{ matrix.march }}