Micro Benchmarks
summary
A practical pair of tools appears: thread instruction counter that counts retired CPU instructions via Puppeteer traces to give a stable, comparable metric (used with npx, reports lowest of runs), and memlab that takes heap snapshots through scripted scenarios to surface leaked objects and representative retainers. Examples show component touches costing ~3×, forced compute jumping 70× then 10× for visible elements, DOM updates 20× (or 4× with fewer ops), and tiny 2–3% differences for hooks versus classes. Tools work today but require Linux and specific Chrome builds and treat instruction counts as comparative, not absolute.
AI-written summary · about
chapters (10)
- 0:00Intro
- 1:05Agenda
- 3:30CPU Retired Instruction Count
- 4:55CPU Instruction Count
- 6:45Thread Instruction Counter
- 9:57AB Testing
- 20:25React hooks
- 24:00Rewrite Chrome
- 27:55Memory Leaks
- 33:40Performance Calendar
related talks
02h 44mMeasure, don’t guess: benchmarking stories from the trenches
27mAnd Benchmarks for All!
21mBenchmarking Like a Boss
02h 41mMicro Benchmarking - The Art of Realizing One is Wrong
59mCounting Nanoseconds Microbenchmarking C++ Code
47mMisnomers About Microbenchmarks
30mPerformance Tuning Secrets
09mNodejs Runtime Performance Tips