JavaScript & Metaperformance
summary
Argues that tuning engines to synthetic tests encourages bad coding patterns and instead creates a new, representative benchmark: JSLint running on itself to measure real JavaScript behavior. Reports concrete numbers — V8 (Chrome) is initially 3–5× slower than Safari and IE10 preview on that workload — and shows the practical result: engine teams deliver a fivefold V8 improvement in about two weeks (Chrome 13 Canary), proving targeted benchmarks change implementation choices.
AI-written summary · about
chapters (7)
- 0:00Start
- 2:18No time for Optimizing
- 6:15Lies, Damn Lies, and Benchmarks
- 7:34Life on the Web
- 8:24The DOM Bottleneck
- 10:46This behavior of developers interacts badly with performance benchmarks. • Benchmarks
- 12:55I designed a new benchmark.
