基准测试规范:benchstat 与性能回归防线
依据单次基准测试结果做出优化决策并不可靠。ns/op 数值受 CPU 频率漂移、后台进程抢占 CPU 核心、GC 时机不固定等环境因素影响,同一份代码两次运行的差异达到数个百分点属于正常现象。单次采样的 ns/op 不具备统计效力,不应作为性能结论的依据。
10 min read
Loading…
Loading…
Quote of the day · November 15关于接受Breaking Bad
You got the job. Do yourself a favor and learn to take yes for an answer.
Notes on Agent runtimes — design decisions laid open: why it is this way, what it costs, and when it stops being true.