1928199 - 7.83% perf_reftest_singletons insert-nth-edge-has-pseudo.html (Windows) regression on Wed October 23 2024

Status: RESOLVED WONTFIX

(Core :: Layout, defect)

Description

[Treeherder Bug Filer]

Perfherder has detected a talos performance regression from push 45ede4b715b4fe8ef8d3866e18d2eb0df0634a26. 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)
8%	perf_reftest_singletons insert-nth-edge-has-pseudo.html	windows11-64-shippable-qr	e10s fission stylo webrender	120.73 -> 130.17

Improvements:

Ratio	Test	Platform	Options	Absolute values (old vs new)
21%	perf_reftest_singletons remove-child-with-lengthy-text.html	linux1804-64-shippable-qr	e10s fission stylo webrender	2.65 -> 2.10
17%	perf_reftest_singletons remove-child-with-lengthy-text.html	linux1804-64-shippable-qr	e10s fission stylo webrender	2.60 -> 2.16

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 42378

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]]

Set release status flags based on info from the regressing bug 1926512

Comment 2

[Jonathan Kew [:jfkthame]]

In general the change in bug 1926512 is expected to make things more efficient, as the cached pointers become simple member accesses rather than lookups into the frame properties. But this does make the actual frame class larger, and so it's possible that this hurts performance (e.g. by impacting allocation patterns) in some cases, depending on the exact patterns of usage. If a testcase does a lot of nsSplittableFrame allocations, but does not actually make any use of the first-continuation/first-in-flow pointers, then bloating the frame class could be a net negative.

Overall, though, the change still looks like a win: the improvements shown in comment 0 are substantially larger than the regression, and the example in bug 1925916 also showed a large improvement (see bug 1925916 comment 15).

So I think we should accept the regression on insert-nth-edge-has-pseudo, as it is outweighed by large improvements elsewhere.

Comment 3

[Donal Meehan [:dmeehan]]

:jfkthame do you want to resolve this as wontfix?

Comment 4

[Jonathan Kew [:jfkthame]]

Yes, I think that makes the most sense here.