1670542 - 544.1% glvideo Mean tick time across 100 ticks: (windows7-32-shippable) regression on push 91f4934e4063a412c9057545dc6707b8f0fc1413 (Wed October 7 2020)

Status: RESOLVED WORKSFORME

(Core :: Graphics: CanvasWebGL, defect, P1)

Description

[Alexandru Ionescu (needinfo me) [:alexandrui]]

Perfherder has detected a talos performance regression from push 91f4934e4063a412c9057545dc6707b8f0fc1413. As author of one of the patches included in that push, we need your help to address this regression.

Regressions:

544% glvideo Mean tick time across 100 ticks: windows7-32-shippable opt e10s stylo 5.29 -> 34.08
530% glvideo Mean tick time across 100 ticks: windows10-64-shippable-qr opt e10s stylo 5.29 -> 33.34
526% glvideo Mean tick time across 100 ticks: windows10-64-shippable opt e10s stylo 5.33 -> 33.35
44% glterrain windows7-32-shippable opt e10s stylo 1.27 -> 1.82
39% glterrain windows10-64-shippable opt e10s stylo 1.25 -> 1.74
7% tp5o windows10-64-shippable-qr opt e10s stylo 240.75 -> 257.33

Improvements:

21% motionmark_webgl 3DGraphics-WebGL windows7-32-shippable opt e10s stylo 6.34 -> 7.68
16% motionmark_webgl 3DGraphics-WebGL windows10-64-shippable opt e10s stylo 6.92 -> 8.00
13% motionmark_webgl 3DGraphics-WebGL windows10-64-shippable-qr opt e10s stylo 6.89 -> 7.81

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.

Comment 1

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

Set release status flags based on info from the regressing bug 1669530

Comment 2

[Kelsey Gilbert [:jgilbert]]

Yep, this is known and expected from what I'm reading.

• Perf in general is a little worse because of buffer-bloat and webgl process scheduling
• Video uploads are no longer low-overhead, so glvideo will be much worse

These are both TODOs but not blockers.

Comment 4

[Alexandru Ionescu (needinfo me) [:alexandrui]]

There's no activity for 11 months here. Any news here?

Comment 5

[Kelsey Gilbert [:jgilbert]]

Today:
glterrain: ~1.38 (only slightly worse than before regression)
glvideo: 4.32 ms (better than before)
motionmark_webgl: ~12.1 fps (much better than before)