Stream: t-compiler/wg-prioritization/alerts

Topic: I-nominated #58368 A big regression in tokio-webpush-simple-

triagebot (Jul 08 2020 at 15:11, on Zulip):

@WG-prioritization/alerts #58368 has been nominated for discussion in T-compiler meeting.

Procedure

Already discussed?
Worth the meeting time?
Add agenda entry:
- Why nominated?
- Assignee?
- Issue? PR? What's the status?
- Summary and important details?

Santiago Pastorino (Jul 08 2020 at 16:40, on Zulip):

Tokio webpush bench regressed 88.5% in this commit.
Nominated to raise awareness of such a big performance regression that didn't have a lot of attention.

Santiago Pastorino (Jul 08 2020 at 16:41, on Zulip):

cc @LeSeulArtichaut thoughts?

Wesley Wiser (Jul 08 2020 at 16:49, on Zulip):

Hmm... this is tricky because it's been so long that the perf.rlo data isn't available any more.

DPC (Jul 08 2020 at 17:07, on Zulip):

is there a way we can re-perf it?

Santiago Pastorino (Jul 08 2020 at 17:38, on Zulip):

Wesley Wiser said:

Hmm... this is tricky because it's been so long that the perf.rlo data isn't available any more.

yeah, so ... can we maybe run benchmarks with current master code and then run again with master + that identified commit reverted?

LeSeulArtichaut (Jul 08 2020 at 17:38, on Zulip):

Should we do perf runs by reverting potential culprits?

Santiago Pastorino (Jul 08 2020 at 17:39, on Zulip):

I have no idea if would be as easy as reverting that commit, let's pretend I have said "revert" between quotes :), I meant, chaging the code to what it looked like before and with it working on current master

Santiago Pastorino (Jul 08 2020 at 17:40, on Zulip):

DPC said:

is there a way we can re-perf it?

I guess in the way I've said, wouldn't be a correct thing to run perf with master and compare that run with the old code because there are a lot of unrelated improvements since then and we would be comparing non sense stuff

Santiago Pastorino (Jul 08 2020 at 17:40, on Zulip):

LeSeulArtichaut said:

Should we do perf runs by reverting potential culprits?

what do you mean?

Santiago Pastorino (Jul 08 2020 at 17:41, on Zulip):

anyway, when I've said thoughts, what I meant was if nomination bullets were more or less correct and summarizes the issue correctly :)

LeSeulArtichaut (Jul 08 2020 at 17:41, on Zulip):

I mean basically what you said at the same time

DPC (Jul 08 2020 at 17:49, on Zulip):

LeSeulArtichaut said:

Should we do perf runs by reverting potential culprits?

yep that's what is done sometimes in such cases

Wesley Wiser (Jul 08 2020 at 19:17, on Zulip):

Yeah, I think that's the best we can do. I just hope reverting that commit isn't too messy.

triagebot (Jul 09 2020 at 14:51, on Zulip):

#58368's nomination has been removed.

Last update: Apr 11 2021 at 19:00UTC