What could cause this sort of spike to occur? Is this a sign the daemon wasn’t “warm” enough (there were 6 warm-ups for each scenario in this case)
Gabriel Feo
02/10/2022, 10:22 PM
By the way, this was running inside a container on a CI machine dedicated to this benchmark at that point in time. The container image was derived from the official Gradle image and no processes are started other than
gradle-profiler
w
wolfs
02/11/2022, 7:29 AM
Garbage collection would be my assumption. Did you capture GC times?
n
no
02/11/2022, 7:59 AM
Could also be thermal throttling.
l
lptr
02/11/2022, 1:37 PM
You can capture GC times by passing
--measure-gc
on the command-line
g
Gabriel Feo
02/22/2022, 2:32 PM
Sorry I missed this. I didn’t look at GC, great point! Thank you all
@no I hope that’s improbable on AWS machines, but that’s a good point for local benchmarks. I hadn’t realized that could be a factor
n
no
02/22/2022, 3:26 PM
Even AWS machines can thermally throttle. We've experienced this at Gradle when running benchmarks. Even when the vendor claimed that thermal throttling was disabled.