Closed Bug 1745475 Opened 2 years ago Closed 2 years ago

432.16 - 4.19% facebook-nav.marketplace LastVisualChange / welcome ContentfulSpeedIndex + 4 more (Linux, OSX) regression on Fri December 3 2021

Categories

(Core :: Graphics: WebRender, defect)

Firefox 97
defect

Tracking

()

RESOLVED INVALID
Tracking Status
firefox-esr91 --- unaffected
firefox95 --- unaffected
firefox96 - wontfix
firefox97 - wontfix

People

(Reporter: alexandrui, Unassigned)

References

(Regression)

Details

(Keywords: perf, perf-alert, regression)

Perfherder has detected a browsertime performance regression from push 356c521826254fbae435b212923b20801d171950. 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)
432% facebook-nav.marketplace LastVisualChange linux1804-64-shippable-qr cold webrender 1,186.67 -> 6,315.00
429% facebook-nav.marketplace LastVisualChange linux1804-64-shippable-qr cold fission webrender 1,195.00 -> 6,326.67
423% facebook-nav.marketplace LastVisualChange macosx1015-64-shippable-qr cold webrender 1,195.00 -> 6,246.67
422% facebook-nav.marketplace LastVisualChange macosx1015-64-shippable-qr cold fission webrender 1,201.67 -> 6,270.00
290% facebook-nav.marketplace LastVisualChange macosx1014-64-shippable-qr cold fission webrender 1,755.00 -> 6,841.67
4% welcome ContentfulSpeedIndex macosx1015-64-shippable-qr cold webrender 1,450.88 -> 1,511.67

Improvements:

Ratio Test Platform Options Absolute values (old vs new)
24% cnn-nav.landing fcp macosx1015-64-shippable-qr cold fission webrender 1,253.04 -> 952.08

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 offending patch(es) will be backed out in accordance with our regression policy.

For more information on performance sheriffing please see our FAQ.

Flags: needinfo?(aosmond)

There should be a way to download videos of these screen caps to see what the change is to help diagnose this.

Unfortunately we don't record the interactive tests yet so there's no way of doing a video comparison.

Set release status flags based on info from the regressing bug 1743761

S3 doesn't feel like the right severity given the magnitude of the regression. Clearing it for re-triage.

Severity: S3 → --

I have been looking at this although I don't have an explanation right now. Backing out the change is difficult because it is so large and sweeping but if I do it, I don't think I will find to time to reland again and we'll let future blob improvement die.

I didn't notice before because we didn't credit bug 1711061 with improving it in the first place.

No longer blocks: gfx-triage
Severity: -- → S3
Status: NEW → RESOLVED
Closed: 2 years ago
Resolution: --- → INVALID
See Also: → 1711061
Flags: needinfo?(aosmond)

Sorry that was unclear. It shows there was no improvement or regression on Windows. On OSX, you can see on Nov 27th, it was the same as it was on Dec 3rd -- the first improved build contained bug 1711061 which introduced a bug (and apparently it is faster if we never redecode images properly):

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=c4b3480996ec9bdbea7040e5cb9e05215d9acb16&tochange=afed7ee7a5dcee09542f3157224a54057527d842

And the first reverted build contains my fix for that in bug 1743761:

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=260e2362bbc3acecc2df9cc1561de8af348a20a2&tochange=356c521826254fbae435b212923b20801d171950

It is very timing sensitive so I could easily see Windows not seeing the problem manifest.

Flags: needinfo?(aosmond)
Has Regression Range: --- → yes
You need to log in before you can comment on or make changes to this bug.