Closed Bug 1920084 Opened 1 month ago Closed 1 month ago

12.57 - 3.16% speedometer3 NewsSite-Nuxt/NavigateToPolitics/Sync / speedometer3 NewsSite-Next/NavigateToUS/total + 23 more (Linux, OSX, Windows) regression on Tue September 17 2024

Categories

(Core :: Layout, defect)

Product:
Core
Component:
Layout
Type:
defect

Tracking

()

Status:
RESOLVED FIXED
Milestone:
132 Branch
Tracking Flags:
Tracking Status
firefox-esr115 --- unaffected
firefox-esr128 --- unaffected
firefox130 --- unaffected
firefox131 --- unaffected
firefox132 --- fixed

People

(Reporter: intermittent-bug-filer, Assigned: TYLin)

References

(Regression)

Details

(Keywords: perf, perf-alert, regression)

Attachments

(1 file)

Bug 1920084 - Revert Bug 1918576 Part 2 to fix performance regression in speedometer3. r?dholbert
48 bytes, text/x-phabricator-request
Details | Review

Perfherder has detected a browsertime performance regression from push 51fd81b5ced0f6cec53b3b475fa50a9a48010848. 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
13% speedometer3 NewsSite-Nuxt/NavigateToPolitics/Sync windows11-64-shippable-qr fission webrender 15.25 -> 17.17 Before/After
11% speedometer3 NewsSite-Nuxt/NavigateToWorld/Sync windows11-64-shippable-qr fission webrender 15.57 -> 17.29 Before/After
9% speedometer3 NewsSite-Nuxt/NavigateToUS/Sync windows11-64-shippable-qr fission webrender 19.42 -> 21.10 Before/After
9% speedometer3 NewsSite-Next/NavigateToPolitics/Sync windows11-64-shippable-qr fission webrender 21.65 -> 23.52 Before/After
7% speedometer3 NewsSite-Next/NavigateToWorld/Sync windows11-64-shippable-qr fission webrender 22.86 -> 24.49 Before/After
6% speedometer3 NewsSite-Nuxt/NavigateToWorld/total linux1804-64-shippable-qr fission webrender 69.31 -> 73.40 Before/After
6% speedometer3 NewsSite-Next/NavigateToUS/Sync windows11-64-shippable-qr fission webrender 25.46 -> 26.94 Before/After
6% speedometer3 NewsSite-Nuxt/NavigateToPolitics/total macosx1015-64-shippable-qr fission webrender 59.32 -> 62.69 Before/After
6% speedometer3 NewsSite-Nuxt/NavigateToPolitics/total windows11-64-shippable-qr fission webrender 34.50 -> 36.40 Before/After
6% speedometer3 NewsSite-Next/NavigateToPolitics/total linux1804-64-shippable-qr fission webrender 74.11 -> 78.19 Before/After
... ... ... ... ... ...
4% speedometer3 NewsSite-Next/NavigateToWorld/total windows11-64-shippable-qr fission webrender 40.88 -> 42.63 Before/After
4% speedometer3 NewsSite-Nuxt/NavigateToUS/total windows11-64-shippable-qr fission webrender 40.66 -> 42.39 Before/After
4% speedometer3 NewsSite-Next/total macosx1015-64-shippable-qr fission webrender 215.63 -> 224.44 Before/After
4% speedometer3 NewsSite-Next/total windows11-64-shippable-qr fission webrender 132.26 -> 137.35 Before/After
3% speedometer3 NewsSite-Next/NavigateToUS/total windows11-64-shippable-qr fission webrender 53.87 -> 55.57 Before/After

Improvements:

Ratio Test Platform Options Absolute values (old vs new) Performance Profiles
3% speedometer3 Editor-TipTap/Long/Sync macosx1015-64-shippable-qr fission webrender 108.73 -> 105.62 Before/After
3% speedometer3 Editor-TipTap/Long/total macosx1015-64-shippable-qr fission webrender 111.24 -> 108.06 Before/After
3% speedometer3 Editor-TipTap/total macosx1015-64-shippable-qr fission webrender 209.66 -> 204.21 Before/After
2% speedometer3 Editor-TipTap/Highlight/Sync macosx1015-64-shippable-qr fission webrender 94.88 -> 92.68 Before/After
2% speedometer3 Editor-TipTap/Highlight/total macosx1015-64-shippable-qr fission webrender 98.43 -> 96.14 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 these tests on try with ./mach try perf --alert 2120

For more information on performance sheriffing please see our FAQ.

Flags: needinfo?(aethanyc)

Set release status flags based on info from the regressing bug 1918576

status-firefox130: --- → unaffected
status-firefox131: --- → unaffected
status-firefox-esr115: --- → unaffected
status-firefox-esr128: --- → unaffected

Bug 1918576 can cause a flex item to recalculate its intrinsic inline size when its block size changes, especially if the intrinsic inline size depends on the block size. The fix ensures correct layout results, and we may have to accept the regression unless we find a smarter way to reduce the work. For example, we could investigate whether NS_FRAME_DESCENDANT_INTRINSIC_ISIZE_DEPENDS_ON_BSIZE is only set when necessary.

Daniel, do you have other ideas?

Flags: needinfo?(aethanyc) → needinfo?(dholbert)

(In reply to Ting-Yu Lin [:TYLin] (PDT, UTC-7) from comment #2)

unless we find a smarter way to reduce the work. For example, we could investigate whether NS_FRAME_DESCENDANT_INTRINSIC_ISIZE_DEPENDS_ON_BSIZE is only set when necessary.

It would probably be worth examining under-what-conditions we're doing more work now, on the particular micro-benchmark here. We should see if the new branches that we're taking are actually buying us any correctness wins in this specific case -- I assume it's not, and maybe we can add some additional checks to skip the work, if there's a coherent testable condition we can extract from this investigation?

Flags: needinfo?(dholbert) → needinfo?(aethanyc)

Daniel, thanks for the direction! Given Firefox 132 will be soft freeze within one week, and I'll have low availability next week. I'll post a patch to revert Bug 1918576 Part 2 to fix this performance regression. As a result, I'll reopen bug 1910290, and we can continue to find a proper solution there.

Flags: needinfo?(aethanyc)
Assignee: nobody → aethanyc
Status: NEW → ASSIGNED
Pushed by aethanyc@gmail.com: https://hg.mozilla.org/integration/autoland/rev/11e366a35b97 Revert Bug 1918576 Part 2 to fix performance regression in speedometer3. r=dholbert,emilio

https://hg.mozilla.org/mozilla-central/rev/11e366a35b97

Status: ASSIGNED → RESOLVED
Closed: 1 month ago
status-firefox132: affected → fixed
Resolution: --- → FIXED
Target Milestone: --- → 132 Branch

(In reply to Pulsebot from comment #6)

Pushed by aethanyc@gmail.com:
https://hg.mozilla.org/integration/autoland/rev/11e366a35b97
Revert Bug 1918576 Part 2 to fix performance regression in speedometer3.
r=dholbert,emilio

Perfherder has detected a browsertime performance change from push 11e366a35b97d14062c7e339ef43b7fd2863b223.

Improvements:

Ratio Test Platform Options Absolute values (old vs new) Performance Profiles
16% twitter largestContentfulPaint windows11-64-shippable-qr bytecode-cached fission warm webrender 632.48 -> 530.62
15% twitter largestContentfulPaint windows11-64-shippable-qr fission warm webrender 636.20 -> 543.95
13% twitter largestContentfulPaint linux1804-64-shippable-qr fission warm webrender 1,028.18 -> 891.52 Before/After
13% speedometer3 NewsSite-Nuxt/NavigateToPolitics/Sync windows11-64-nightlyasrelease-qr fission webrender 16.98 -> 14.82
12% speedometer3 NewsSite-Nuxt/NavigateToPolitics/Sync windows11-64-shippable-qr fission webrender 17.34 -> 15.30 Before/After
... ... ... ... ... ...
2% speedometer3 NewsSite-Next/NavigateToUS/total macosx1400-64-shippable-qr fission webrender 28.49 -> 27.84 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.

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 2208

For more information on performance sheriffing please see our FAQ.

You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: