Shimin Guo

/remove-lifecycle stale

Also ran into the same issue. @hardc0der did you somehow resolve it?

From https://perf.wiki.kernel.org/index.php/Tutorial#Event-based_sampling_overview: > The period is expressed as the number of occurrences of an event, not the number of timer ticks. And later on the same page: > ... the...

I take that to mean that this sample was taken after that many cycle events fired, not that the IP stayed there for that long.

I guess we should weight the samples by this amount. I'm going to check how `perf report` does it.

Yeah `perf report` weight the samples by period. The code is https://github.com/torvalds/linux/blob/5925fa68fe8244651b3f78a88c4af99190a88f0d/tools/perf/ui/hist.c#L36-L40

Yeah I'm getting astronomical numbers that clearly don't make sense in the `period` field as well when I'm using the `cycles` event. I don't know if it's a kernel issue...

Yeah I think that's fair. Maybe recommend setting `-c` instead of `-F` when using `cycles` to get constant periods.

Looks like it's a simple matter of getting the model jsons from the go sdk, and regenerate the Haskell code from them. Since the contributor guide says not to submit...

/remove-lifecycle stale On Tue, Dec 14, 2021, 5:00 AM Kubernetes Triage Robot < ***@***.***> wrote: > The Kubernetes project currently lacks enough contributors to adequately > respond to all issues...