A Github Action that runs your Android microbenchmarks in Firebase Test Labs and optionally:
- updates a GitHub issue with the latest benchmark results (see here for a sample issue)
- publishes the metrics to Datadog (see here for a sample dashboard)
See the action in action (ha ha!) at https://github.com/martinbonnin/run-benchmarks-sample
Note: Macrobenchmarks are not supported just yet
on:
schedule:
# Run every night
- cron: '0 3 * * *'
jobs:
benchmarks:
runs-on: ubuntu-20.04
steps:
- uses: actions/checkout@v3
- uses: actions/setup-java@v3
with:
distribution: 'temurin'
java-version: '11'
- uses: gradle/[email protected]
- run: |
# Build the benchmark apks
./gradlew :microbenchmark:packageReleaseAndroidTest :app:assembleRelease
- name: microbenchmarks
uses: martinbonnin/run-benchmarks@main
with:
google_services_json: ${{ secrets.GOOGLE_SERVICES_JSON }}
app_apk: 'app/build/outputs/apk/release/app-release.apk'
test_apk: 'microbenchmark/build/outputs/apk/androidTest/release/microbenchmark-release-androidTest.apk'
device_model: 'redfin,locale=en,orientation=portrait'
# Optional, upload to datadog
dd_api_key: ${{ secrets.DD_API_KEY }}
dd_metric_prefix: 'android.benchmark'
# Optional, create a dashboard issue that publishes the latest results
github_token: ${{ github.token }}
dd_dashboard_url: 'https://p.datadoghq.com/sb/'If you specify dd_api_key, the metrics are uploaded to Datadog automatically. This action publishes 2 metrics:
- ${dd_metric_prefix}.allocs: the median number of allocations per test
- ${dd_metric_prefix}.nanos: the median number of nanosecond per test
Each metric is tagged with the class of the test and the name of the test. For an example:
class:com.example.BenchmarkTesttest:parseSomething