1920878 - 9.33 - 7.77% speedometer3 cpuTime / speedometer3 cpuTime (Linux) regression on Fri September 20 2024

Status: RESOLVED WONTFIX

(Firefox :: Address Bar, defect)

Description

[Treeherder Bug Filer]

Perfherder has detected a browsertime performance regression from push a90481dcd451b1b03055be15830b257ce3f4872a. As author of one of the patches included in that push, we need your help to address this regression.

Regressions:

Ratio	Test	Platform	Options	Absolute values (old vs new)	Performance Profiles
9%	speedometer3 cpuTime	linux1804-64-nightlyasrelease-qr	fission webrender	635.83 -> 695.17	Before/After
8%	speedometer3 cpuTime	linux1804-64-nightlyasrelease-qr	fission webrender	638.43 -> 688.05	Before/After

Details of the alert can be found in the alert summary, including links to graphs and comparisons for each of the affected tests. Please follow our guide to handling regression bugs and let us know your plans within 3 business days, or the patch(es) may be backed out in accordance with our regression policy.

If you need the profiling jobs you can trigger them yourself from treeherder job view or ask a sheriff to do that for you.

You can run all of these tests on try with ./mach try perf --alert 2220

The following documentation link provides more information about this command.

For more information on performance sheriffing please see our FAQ.

If you have any questions, please do not hesitate to reach out to [e-mail].

Comment 1

[BugBot [:suhaib / :marco/ :calixte]]

Set release status flags based on info from the regressing bug 1914390

Comment 2

[Marco Bonardo [:mak]]

If I compare the profiles, the before profile has work done in SearchModeSwitcher, while the after profile doesn't have that, so we're effectively doing less in urlbar. This is cpuTime, so I'm wondering if something entered the measurement bracket (I see a few markers happening a bit earlier).

Comment 3

[Daisuke Akatsuka (:daisuke)]

Thank you, Marco!
It might not affect the performance much, but I realized that the switcher continued to observe the pref change even after the owner window was closed. I made a patch that avoids it.
https://treeherder.mozilla.org/jobs?repo=try&revision=ce21b2774ff9a0bfdc922e674ce879afb6727c89

Comment 4

[Daisuke Akatsuka (:daisuke)]

(Perhaps, it is the first time to use ./mach try perf :)

Comment 5

[Daisuke Akatsuka (:daisuke)]

https://treeherder.mozilla.org/jobs?repo=try&revision=ce21b2774ff9a0bfdc922e674ce879afb6727c89
vs
https://treeherder.mozilla.org/jobs?repo=try&revision=37858c6f2d7a262bde180541c41ef7e3fe11b16c
??

Comment 6

[Daisuke Akatsuka (:daisuke)]

2% better?
https://treeherder.mozilla.org/perfherder/compare?originalProject=try&originalRevision=ce21b2774ff9a0bfdc922e674ce879afb6727c89&newProject=try&newRevision=37858c6f2d7a262bde180541c41ef7e3fe11b16c&framework=13&page=1

Comment 7

[Marco Bonardo [:mak]]

(In reply to Daisuke Akatsuka (:daisuke) from comment #3)

It might not affect the performance much, but I realized that the switcher continued to observe the pref change even after the owner window was closed.

That's strange, we use weak refs for those observers. It could surely be short lived, but I don't expect that to make a difference.

Comment 8

[BugBot [:suhaib / :marco/ :calixte]]

Set release status flags based on info from the regressing bug 1914390

Comment 9

[James Teow [:jteow]]

Hi Florian, would you be able to give us some guidance on what the issue could be based on the profiles? On our side, we thought we would be doing less work so we're not sure why we're seeing this performance regression.

Comment 10

[Greg Mierzwinski [:sparky]]

This alert can be closed as WONTFIX. The cpuTime metric was only measuring the pageload of the test. The name of it recently changed to cpuTimePageload to reflect that: https://treeherder.mozilla.org/perfherder/graphs?series=autoland,212808,1,13&selected=212808,112641058

In bug 1917662, a cpuTimeSupport metric was added for the measurement of the amount of milliseconds used by processes on the CPU to capture the usage during the actual sp3 benchmark test: https://treeherder.mozilla.org/perfherder/graphs?series=autoland,212809,1,13&selected=212809,112641059

Feel free to reopen if you think that an investigation should continue into this change. Something to note is that the cpuTimePageload is back to the value it was at before the regression took place.