And Benchmarks for All!
summary
Framework performance needs concrete evidence not platitudes so the talk lays out actual benchmarks that profile DOM and JavaScript execution, uses benchmark.js for accurate timers, runs Chrome Canary with V8 tracing and parses output with IR hydra, and builds a tiny harness called perform to run the same benchmark against master and a tweak. Results include surprising effects (V8 inlining and hidden classes kills work, inline caching gives ~20x hot-path speed) and the demo comparison works end to end.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:56Why is benchmarking hard
- 2:17What is compiler
- 3:27Constant Propagation
- 6:02Inlining
- 6:42Why do I bother
- 8:13Compilers
- 13:21Benchmarking Patterns
- 16:29Benchmark Jess
- 17:43Demo
- 19:08Metadata
- 20:38Deza
