Skip to content

Latest commit

 

History

History
101 lines (66 loc) · 3.03 KB

README.md

File metadata and controls

101 lines (66 loc) · 3.03 KB

Benchmark

Running the benchmarks

First, build the benchmarks:

yarn build:performance

Then run the benchmarks:

yarn test:performance

Individual benchmark results are saved in *.tachometer.results.json files. A summary table is printed to console and saved as results.md and results.html.

To recreate the summary tables without re-running the full test suite, ensure the JSON files exist do:

cd packages/@lwc/perf-benchmarks
yarn test:format

To run an individual benchmark, do:

cd packages/@lwc/perf-benchmarks
yarn tach --config dist/__benchmarks__/path/to/tachometer.json

Tip

Running all of the benchmarks will take multiple hours. Let it run overnight!

Manual testing

When the benchmark is not working, the best way to debug it locally is to load its HTML file and inspect it in the Chrome DevTools. After building, you can run:

cd packages/@lwc/perf-benchmarks
yarn tach --manual --config dist/__benchmarks__/path/to/tachometer.json

This will print out the URLs you can use to test manually.

When analyzing with the Chrome DevTools Performance tab, pay special attention to the following performance measures in the "Timing" section:

  • benchmark-before
  • benchmark-run
  • benchmark-after

benchmark-run is what's actually measured, whereas the -before and -after measures are just the setup and teardown code.

Testing other branches

By default, the benchmark will compare the local code against the latest master branch from the salesforce/lwc repo. To test against another branch or commit, use the following environment variables when running yarn build:performance:

BENCHMARK_REPO=https://example.com/repo.git \
  BENCHMARK_REF=branchOrTagOrCommit \
  yarn build:performance

You can also use these environment variables to adjust the default benchmark settings:

BENCHMARK_SAMPLE_SIZE=50
BENCHMARK_AUTO_SAMPLE_CONDITIONS=25%
BENCHMARK_TIMEOUT=5
BENCHMARK_CPU_THROTTLING_RATE=4

See the Tachometer documentation for details on what these mean.

For BENCHMARK_CPU_THROTTLING_RATE, some tests have a built-in cpuThrottlingRate which is exported as a const. The environment variable will override this.

You can run a smoke test (to confirm the benchmark tests are working) using:

BENCHMARK_SMOKE_TEST=1

If anything gets messed up when comparing to the other branch, add the --force-clean-npm-install flag when running tach.

Also note that, due to how NX does caching, any environment variables you pass to build:performance should also be passed to test:performance. Otherwise, test:performance will use whatever environment variables you give it (or don't give it) to re-run build:performance.

Best benchmarks

This package also supports Best as a benchmark runner. To run the Best tests locally:

yarn test:performance:best

Or for CI:

yarn test:performance:best:ci

Tip

Best tests only take a few dozen minutes. Take a break!