chromium-notes

Benchmarks

System health

Below are some examples for running the system health bench

• memory_desktop
  • tools/perf/run_benchmark system_health.memory_desktop --browser debug --story-filter load:games:miniclip:2018 --reset-results
• common_desktop
  • tools/perf/run_benchmark system_health.common_desktop --browser debug --story-filter load:news:reddit:2018 --reset-results

You typically want to only add the --reset-results flag when you start a new series of runs that you’d like to compare side-by-side on the results page that is emitted after the benchmark runs.

Pausing benchmarks

Running benchmarks locally is largely automated, but sometimes it is useful to pause a benchmark, open DevTools, and manually inspect things in the DOM, the console, or the networking tab for example. A hacky way to do this is to open third_party/catapult/telemetry/telemetry/internal/story_runner.py, and add the following lines after L141:

print("State RunStory")
time.sleep(120000)