The Modern Prometheus
summary
Argues that common benchmarks mislead and demonstrates concrete failures: exceptions are ~20× slower than a hand‑written path, v1.0 crashes when job queue exceeds the limit, shows three bit‑count implementations including slice an integer into four parts and count set bits per part with measurable variance, compares string‑replace strategies and marshaling approaches, and reports surprising cross‑language results (one Rust build is reported ~36× slower in a run), concluding always measure real workloads.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 3:48Benchmarks and exceptions
- 7:54Performance considerations
- 11:05Bit counting algorithm
- 15:59Optimizing bit operations
- 23:28String processing performance
- 26:35Benchmarking pitfalls
- 30:51Stream processing
- 33:37Language limitations
- 36:51Performance without code changes
- 38:03Language design impact
- 39:24Closing remarks
related talks
02h 44mMeasure, don’t guess: benchmarking stories from the trenches
02h 41mMicro Benchmarking - The Art of Realizing One is Wrong
54mPerformance tests and other lies
43mStop Guessing and Start Measuring - Benchmarking in Practice
38mYour benchmark may not guide a real application performance
31mMaking it Fast: The Power of Benchmarking
39mMicro Benchmarks
45mPerformance Benchmarking with Celero