1467169 - tsvgx regression with WebRender from SVG-as-a-reference-frame

Status: RESOLVED WONTFIX

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

Description

[Markus Stange [:mstange]]

This is the WebRender-specific part of bug 1464046.


Talos has detected a Firefox performance regression from push:

https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?changeset=133a13c44abedac2e448d315a32068ce1a5568f4

Regressions:

 17%  tsvgx linux64-qr opt e10s stylo     328.21 -> 385.57
  8%  tsvg_static linux64-qr opt e10s stylo54.73 -> 59.19

You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=13394


In bug 1464046 comment 6, Jeff reported that this regression only manifests when using software OpenGL.

Comment 1

[Mike Taylor [:miketaylr]]

David, could you help find someone to own, or investigate this please?

Comment 2

[Markus Stange [:mstange]]

This will be worked on as part of bug 1416652, which aims to eliminate the WebRender vs non-WebRender regression on this test. The specific regression that this bug was filed about does not affect 62 because it's WebRender-only and WebRender is not turned on by default yet.

Comment 3

[Mike Taylor [:miketaylr]]

Ah, thanks Markus. Let's mark it as disabled for 62 then.

Comment 4

[Kartikaya Gupta (email:[e-mail])]

By itself this regression is not a release blocker. Bug 1416652 is the release blocker, and fixing this is one of the possible ways to address that, but it's possible we can address that in other ways as well.

Comment 5

[Maire Reavy [:mreavy]]

Moving to stage-wr-next because (as Kats says) it doesn't block release.  We can move it back if we decide we need it to fix 1416652

Comment 6

[Jeff Muizelaar [:jrmuizel]]

tsvgx performance is now totally different than it was 8 months ago.