1874897 - 4774.18 - 26.36% perf_reftest bidi-resolution-1.html / perf_reftest + 14 more (Linux, OSX, Windows) regression on Thu January 11 2024

Status: RESOLVED FIXED

(Core :: Layout, defect)

Description

[Andrej (:aglavic)]

Perfherder has detected a talos performance regression from push dfcfa98a7e21b51e5559d4bee58d465a40576f31. 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)
4774%	perf_reftest bidi-resolution-1.html	windows10-64-shippable-qr	e10s fission stylo webrender	5.81 -> 283.19
3372%	perf_reftest bidi-resolution-1.html	windows10-64-shippable-qr	e10s fission stylo webrender-sw	6.52 -> 226.27
3337%	perf_reftest bidi-resolution-1.html	linux1804-64-shippable-qr	e10s fission stylo webrender-sw	6.57 -> 225.80
3330%	perf_reftest bidi-resolution-1.html	linux1804-64-shippable-qr	e10s fission stylo webrender	6.82 -> 233.72
3202%	perf_reftest bidi-resolution-1.html	macosx1015-64-shippable-qr	e10s fission stylo webrender	8.14 -> 268.70
3023%	perf_reftest bidi-resolution-1.html	macosx1015-64-shippable-qr	e10s fission stylo webrender-sw	8.24 -> 257.47
1176%	perf_reftest_singletons bidi-resolution-1.html	windows10-64-shippable-qr	e10s fission stylo webrender	123.18 -> 1,571.35
1154%	perf_reftest_singletons bidi-resolution-1.html	windows10-64-shippable-qr	e10s fission stylo webrender	123.68 -> 1,550.48
1109%	perf_reftest_singletons bidi-resolution-1.html	macosx1015-64-shippable-qr	e10s fission stylo webrender	132.26 -> 1,598.52
937%	perf_reftest_singletons bidi-resolution-1.html	linux1804-64-shippable-qr	e10s fission stylo webrender	183.87 -> 1,906.34
...	...	...	...	...
31%	perf_reftest	linux1804-64-shippable-qr	e10s fission stylo webrender	1.59 -> 2.08
30%	perf_reftest	macosx1015-64-shippable-qr	e10s fission stylo webrender-sw	1.47 -> 1.90
30%	perf_reftest	windows10-64-shippable-qr	e10s fission stylo webrender-sw	1.39 -> 1.81
28%	perf_reftest	linux1804-64-shippable-qr	e10s fission stylo webrender-sw	1.53 -> 1.96
26%	perf_reftest	macosx1015-64-shippable-qr	e10s fission stylo webrender	1.49 -> 1.89

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 these tests on try with ./mach try perf --alert 41006

For more information on performance sheriffing please see our FAQ.

Comment 1

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

Set release status flags based on info from the regressing bug 1873530

Comment 2

[Ting-Yu Lin [:TYLin] (PDT, UTC-7)]

This is a serious regression. I'm investigating.

Comment 3

[Ting-Yu Lin [:TYLin] (PDT, UTC-7)]

Bug 1873530 Part 3 unified the SetNextInFlow and SetPrevInFlow operation. In the new nsTextFrame::SetNextInFlow, it clears the old next-continuations' first-continuation cache [1]. Although it is functionally correct , it however does unnecessarily cache update if the caller wants to insert a new frame between a existing link in nsContinuingTextFrame::Init() [2]. To make it concrete, if we are going to insert a new frame C between the existing link A and B, after A->SetNextInFlow(C), the first-continuation cache in B as well as the cache in its next continuations will be cleared. Then we call C->SetNextContinuation(B), the cache in B will be update again. However, it's OK to not update the cache, because ultimately B is still in the chain and the cached value won't change.

Given Firefox 123 are close to soft freeze, I'll upload a patch to revert Bug 1873530 Part 3 to fix the performance regression.

[1] https://searchfox.org/mozilla-central/rev/2fa4750e87d49ec7ad19fa1da167a03ef6d5e954/layout/generic/nsTextFrame.cpp#4405-4412
[2] https://searchfox.org/mozilla-central/rev/2fa4750e87d49ec7ad19fa1da167a03ef6d5e954/layout/generic/nsTextFrame.cpp#4197,4227

Comment 4

[Ting-Yu Lin [:TYLin] (PDT, UTC-7)]

Attachment: Bug 1874897 - Revert Bug 1873530 Part 3 to fix performance regression. r?#layout

This reverts commit https://hg.mozilla.org/mozilla-central/rev/dfcfa98a7e21
without any additional modification.

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/6717f7be99be
Revert Bug 1873530 Part 3 to fix performance regression. r=dholbert

Comment 6

[Norisz Fay [:noriszfay]]

https://hg.mozilla.org/mozilla-central/rev/6717f7be99be

Comment 7

[Andrej (:aglavic)]

(In reply to Pulsebot from comment #5)

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/6717f7be99be
Revert Bug 1873530 Part 3 to fix performance regression. r=dholbert

Just updating you that your reversion of part 3 has registered as an improvement in our system

== Change summary for alert #41110 (as of Tue, 23 Jan 2024 06:27:49 GMT) ==

Improvements:

Ratio	Test	Platform	Options	Absolute values (old vs new)
98%	perf_reftest bidi-resolution-1.html	windows10-64-shippable-qr	e10s fission stylo webrender-sw	227.11 -> 5.16
98%	perf_reftest bidi-resolution-1.html	windows10-64-shippable-qr	e10s fission stylo webrender	256.33 -> 6.00
97%	perf_reftest bidi-resolution-1.html	linux1804-64-shippable-qr	e10s fission stylo webrender	225.38 -> 6.09
97%	perf_reftest bidi-resolution-1.html	macosx1015-64-shippable-qr	e10s fission stylo webrender	260.06 -> 7.11
97%	perf_reftest bidi-resolution-1.html	linux1804-64-shippable-qr	e10s fission stylo webrender-sw	221.03 -> 7.62
...	...	...	...	...
22%	perf_reftest	macosx1015-64-shippable-qr	e10s fission stylo webrender-sw	1.97 -> 1.53

For up to date results, see: https://treeherder.mozilla.org/perfherder/alerts?id=41110