%0 Journal Article %1 Barrett:2017:VMW %A Barrett, Edd %A Bolz-Tereick, Carl Friedrich %A Killick, Rebecca %A Mount, Sarah %A Tratt, Laurence %D 2017 %I ACM %J Proc. ACM Program. Lang. %K Compiler DynamicCompilation JIT Warmup %N OOPSLA %P 52:1--52:27 %R 10.1145/3133876 %T Virtual Machine Warmup Blows Hot and Cold %V 1 %X Virtual Machines (VMs) with Just-In-Time (JIT) compilers are traditionally thought to execute programs in two phases: the initial warmup phase determines which parts of a program would most benefit from dynamic compilation, before JIT compiling those parts into machine code; subsequently the program is said to be at a steady state of peak performance. Measurement methodologies almost always discard data collected during the warmup phase such that reported measurements focus entirely on peak performance. We introduce a fully automated statistical approach, based on changepoint analysis, which allows us to determine if a program has reached a steady state and, if so, whether that represents peak performance or not. Using this, we show that even when run in the most controlled of circumstances, small, deterministic, widely studied microbenchmarks often fail to reach a steady state of peak performance on a variety of common VMs. Repeating our experiment on 3 different machines, we found that at most 43.5% of <VM, Benchmark> pairs consistently reach a steady state of peak performance.

@article{Barrett:2017:VMW, abstract = {Virtual Machines (VMs) with Just-In-Time (JIT) compilers are traditionally thought to execute programs in two phases: the initial warmup phase determines which parts of a program would most benefit from dynamic compilation, before JIT compiling those parts into machine code; subsequently the program is said to be at a steady state of peak performance. Measurement methodologies almost always discard data collected during the warmup phase such that reported measurements focus entirely on peak performance. We introduce a fully automated statistical approach, based on changepoint analysis, which allows us to determine if a program has reached a steady state and, if so, whether that represents peak performance or not. Using this, we show that even when run in the most controlled of circumstances, small, deterministic, widely studied microbenchmarks often fail to reach a steady state of peak performance on a variety of common VMs. Repeating our experiment on 3 different machines, we found that at most 43.5% of <VM, Benchmark> pairs consistently reach a steady state of peak performance.}, acmid = {3133876}, added-at = {2018-05-11T20:29:44.000+0200}, articleno = {52}, author = {Barrett, Edd and Bolz-Tereick, Carl Friedrich and Killick, Rebecca and Mount, Sarah and Tratt, Laurence}, biburl = {https://www.bibsonomy.org/bibtex/2f4010f403abacd3c6826ffeef4dd11e4/gron}, description = {Virtual machine warmup blows hot and cold}, doi = {10.1145/3133876}, interhash = {14e8c46a45aef0645a3ee6b70dfe0e2b}, intrahash = {f4010f403abacd3c6826ffeef4dd11e4}, issn = {2475-1421}, issue_date = {October 2017}, journal = {Proc. ACM Program. Lang.}, keywords = {Compiler DynamicCompilation JIT Warmup}, month = oct, number = {OOPSLA}, numpages = {27}, pages = {52:1--52:27}, publisher = {ACM}, timestamp = {2018-05-11T20:29:44.000+0200}, title = {{Virtual Machine Warmup Blows Hot and Cold}}, volume = 1, year = 2017 }