Using Benchmarks for Tracking Regressions and Improving UI performance
summary
Benchmarks measure create/bind/layout times for Litho/views, run hourly on a device farm and emit JSON for CI; a monitored header goes from 10 → 14 ms (40% regression), the system runs an automated bisect finds offending commit and assigns a regression task to the commit author. Tests run many iterations and trials on real devices, CPU locked and disk mounted to RAM with local network proxy to reduce noise; results are directional, used for triage and to trigger reverts or optimizations.
AI-written summary · about
chapters (8)
- 0:00Introduction and motivation
- 4:12Why UI performance matters
- 10:20Automating performance tracking
- 14:41Writing benchmark tests
- 23:30Using benchmarks at Facebook
- 29:36Making benchmarks reliable
- 36:24Real-world benchmark insights
- 40:12Q&A session
related talks
37mTracking Regressions and Improving UI Performance using Benchmarking Approach
36mFighting Regressions with Benchmarks in CI
29mHow to Performance
44mAndroid Benchmarking and other stories
33mAndroid Benchmarking and other stories
27mAnd Benchmarks for All!
39mMicro Benchmarks
17mWhat I Learned Benchmarking React