perf triage report construction
Meeting proposal info
- Title: Revisit perf triage process
- Type: it's complicated
Summary
The perf triage process today is largely one person (currently either @Mark-Simulacrum or @rylev If I understand correctly) following a script to review performance changes for the compiler and identify which PR's led to those changes.
One problem with this is that it seems like it might be more arduous than it needs to be.
Also, I'd like to figure out a process that is amenable to distribution+parallelization+delegation.
Finally, rylev noted "(we don't have a good process for ensuring the perf triage is updated as progress is made on addressing issues. We sometimes go back and update the report, but I did not do that this week)." It would be good to figure out if there's some way to dynamically update the report as things shift.
(Its possible the answer here is that we to invest in the perf dashboard itself, to improve it to the point where we don't need to create separate perf triage reports...)
About this issue
This issue corresponds to a meeting proposal for the compiler team steering meeting. It corresponds to a possible topic of discussion. You can read more about the steering meeting procedure here.
Comment policy
These issues are meant to be used as an "announcements channel" regarding the proposal, and not as a place to discuss the technical details. Feel free to subscribe to updates. We'll post comments when reviewing the proposal in meetings or making a scheduling decision. In the meantime, if you have questions or ideas, ping the proposers on Zulip (or elsewhere).
@rylev has a document describing status quo and some ideas for changes; document here: https://hackmd.io/@ryanlevick/Bk1Hidcbu discussed here: https://zulip-archive.rust-lang.org/247081tcompilerperformance/70522Theshorttermperformanceprocessreview.html
This meeting has been scheduled for April 16th at 10am ET (14:00 UTC).
We ended up holding this meeting on 2021-04-30, due to scheduling conflicts.
We discussed this doc: https://hackmd.io/@ryanlevick/Bk1Hidcbu
The public zulip archive will serve as the official record for the meeting: https://rust-lang.zulipchat.com/#narrow/stream/238009-t-compiler.2Fmeetings/topic/.5Bsteering.20meeting.5D.202021-04-30.20perf.20triage.2C.20compiler-team.23400/near/236837851