This is our paper appearing at OOPSLA this year.
TLDR: Programming languages don’t warmup as we might expect.
Pretty unpleasant results - often-inconsistent behavior makes it hard to even define “performance after warmup”, and lots of measurements end up finding that the “steady state” is either unsteady or worse than the startup behavior.
Nice, and quite thorough, work!
I feel like these results are really really rather important. Very good work!
This seems really long for a conference paper. Is it a longer version?
The Appendix was not included in the published version, but instead supplied in the supplementary material archive.