Accessibility support can have a negative impact on performance, so it is important to test for regressions and to improve performance over time. This can be done with Chromium's performance testing framework, Telemetry.
The metrics for accessibility are defined in third_party/catapult/tracing/tracing/metrics/accessibility_metric.html.
A story is a url to load and actions to take on that page.
Accessibility stories are defined in tools/perf/page_sets/system_health/accessibility_stories.py. If a particular web page seems slow, you may add a new story here to track its performance.
To run all accessibility stories locally using the currently installed Canary browser:
tools/perf/run_benchmark system_health.common_desktop \ --story-filter="accessibility.*" \ --browser canary
To run the same set of tests on your own compiled version of Chrome:
tools/perf/run_benchmark system_health.common_desktop \ --story-filter="accessibility.*" \ --browser=exact \ --browser-executable=out/Release/chrome
For more information, see the Telemetry documentation or command-line help for tools/perf/run_benchmark.
You should capture web page replay information whenever you add or modify a story. To do so, see here.
Blink perf tests are microbenchmarks that track the performance of a small function or operation in isolation. You can find these tests in third_party/blink/perf_tests/accessibility/.
To run these tests locally on your own compiled Chrome:
tools/perf/run_benchmark blink_perf.accessibility \ --browser=exact \ --browser-executable=out/Release/chrome
The results can be found at https://chromeperf.appspot.com.
More information on analytics and performance, including some additional Google-internal resources, can be found at go/chrome-a11y/resources/analytics-performance.