“Here is an example of more rigorous research: Virtual Machine Warmup Blows Hot and Cold”
Virtual Machine Warmup Blows Hot and Cold
统计分析显示许多 JIT 虚拟机基准并不能稳定达到峰值性能
传统观点认为带 JIT 编译器的虚拟机分两个阶段执行程序:先预热确定哪些部分值得动态编译,之后进入峰值性能的稳态,而测量方法几乎总是丢弃预热阶段的数据。作者提出基于变点分析的全自动统计方法,判断程序是否达到稳态、该稳态是否为峰值性能。结果表明,即使在最受控的条件下,小而确定、被广泛研究的微基准在多种常见虚拟机上也常常达不到峰值稳态;在 3 台不同机器上重复实验,最多只有 43.5% 的 <VM, 基准> 组合能稳定达到峰值稳态。
HN 上的人把它当作更严谨的基准测试研究,用来说明 JIT 的稳态性能难以预测、也难以可靠测量,不能保证获得稳定的好性能。
HN 书评
推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及“Virtual Machine Warmup Blows Hot and Cold is a good exploration of why this just doesn't work out”
“To see that just look at a program, and predict how long the warm-up will take. Whoops... unknown [1].”
“Indeed the performance benefits of JITing have been difficult to measure: Virtual Machine Warmup Blows Hot and Cold”
“The snapshots of the paper we released are at Virtual Machine Warmup Blows Hot and Cold”
“JITs are a great magic trick but it's nowhere near guaranteed you'll get good steady performance out of one, especially when the workload is wide not narrow.”
- Virtual Machine Warmup Blows Hot and Cold ↗ingve · 2016-02-02 · 5 分