1405327 - 5.62 - 9.35% Strings PerfIsASCIIExample3 / Strings PerfIsASCIIHundred (macosx64-nightly) regression on push 04e03b0726f97c5f89db4e08e1029147d4753063 (Thu Sep 28 2017)

Status: RESOLVED WORKSFORME

(Core :: CSS Parsing and Computation, defect)

Description

[Ionuț Goldan [:igoldan]]

We have detected a platform microbenchmarks regression from push:

https://hg.mozilla.org/releases/mozilla-beta/pushloghtml?changeset=04e03b0726f97c5f89db4e08e1029147d4753063

As author of one of the patches included in that push, we need your help to address this regression.

Regressions:

  9%  Strings PerfIsASCIIHundred macosx64-nightly opt      1,922.00 -> 2,101.67
  6%  Strings PerfIsASCIIExample3 macosx64-nightly opt     8,232.96 -> 8,695.46


You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=9740

On the page above you can see an alert for each affected platform as well as a link to a graph showing the history of scores for this test. There is also a link to a treeherder page showing the jobs in a pushlog format.

To learn more about the regressing test(s), please see: https://developer.mozilla.org/en-US/docs/Mozilla/Performance/Automated_Performance_Testing_and_Sheriffing/Platform_Microbenchmarks

Comment 1

[Ionuț Goldan [:igoldan]]

This regression appeared on Beta after an uplift. As a this patch included multiple bugs, bisection is required to point the right changeset. Until then... From the bulk of changes, looks like bug 1403282 may be related to this.

Manish, can you provide feedback about this?

Comment 2

[Ionuț Goldan [:igoldan]]

ni?ing others to make sure we identify this quickly.

Comment 3

[Emilio Cobos Álvarez (:emilio)]

I'm moderately sure this is noise, or just the optimizer changing code depending on one of the patches. No patch in that range seems to touch xpcom/ at all.

Comment 4

[Manish Goregaokar [:manishearth]]

I agree with emilio, those patches change how we parse attr(), which is rare enough that it shouldn't affect perf.

Comment 5

[Michael Comella (:mcomella) [NI reported issues only: ex-Mozilla]]

I made Android front-end changes, I don't think I affected this.

Comment 6

[Mike Kaply [:mkaply]]

My changes were around webextension notification styles. Shouldn't be me.

Comment 7

[Jet Villegas (inactive)]

(In reply to Emilio Cobos Álvarez [:emilio] from comment #3)
> I'm moderately sure this is noise, or just the optimizer changing code
> depending on one of the patches. No patch in that range seems to touch
> xpcom/ at all.

Recorded times after 9/28 seem to agree: 
https://treeherder.mozilla.org/perf.html#/graphs?timerange=7776000&series=%5Bmozilla-beta,21fb1e480d6afbd6500dda867d808d1e57d6060b,1%5D&selected=%5Bmozilla-beta,21fb1e480d6afbd6500dda867d808d1e57d6060b%5D

Comment 8

[Ionuț Goldan [:igoldan]]

Thanks guys, for helping me sort this out in such quick time!