When using WebRender with proprietary Nvidia drivers, dragging tabs is much slower and more sluggish than when not using it
Categories
(Core :: Graphics: WebRender, defect, P3)
Tracking
()
Tracking | Status | |
---|---|---|
firefox81 | --- | disabled |
People
(Reporter: u658943, Unassigned)
References
(Blocks 1 open bug)
Details
(Keywords: nightly-community, perf)
User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:81.0) Gecko/20100101 Firefox/81.0
Steps to reproduce:
- Open firefox (with webrender enabled)
- Open a second tab
- Drag the tab back and forth before and after the other tab
Actual results:
When webrender is enabled the tab dragging is much more sluggish and slow when compared to not using webrender
Expected results:
The tab should have dragged smoothly, like it does when webrender is disabled
Video showcasing the issue: https://www.reddit.com/r/firefox/comments/icaf2r/firefox_lags_when_dragging_tabs_in_webrender/
Performance profile with webrender enabled: https://share.firefox.dev/2EgWSgS
Performance profile with webrender disabled: https://share.firefox.dev/34dHA7I
Comment 1•4 years ago
|
||
KDE/X11 (compositor disabled), Debian Testing, Nvidia GTX1060, proprietary driver 440.100
mozregression --launch 2020-08-18 --pref gfx.webrender.all:true
mozregression --launch 2020-08-18 --pref gfx.webrender.force-disabled:true
Confirmed, it happens only with WebRender and the proprietary Nvidia driver.
(The drag symbol looks like a separate window: Maybe it's causing the performance problem in the underlying main window?)
Updated•4 years ago
|
Updated•4 years ago
|
Comment 3•4 years ago
|
||
Has this been observed/confirmed on non-nvidia? The proprietary nvidia driver is known to have all kind of bugs like this (just look at the Mutter bug tracker...). If it only happens on nvidia, can we add that to the title?
Comment 4•4 years ago
|
||
So far I try to put bugs that possibly only occur on proprietary Nvidia drivers under the wr-nv-linux metabug (bug 1535716).
Comment 5•3 years ago
|
||
MOZ_DISABLE_CONTENT_SANDBOX=1 mozregression --find-fix --bad 2020-08-18 --good 2021-09-15 --pref gfx.webrender.all:true gfx.x11-egl.force-disabled:true -a about:support -a about:support -a about:support -a about:support -a about:support
15:57.48 INFO: First good revision: f8febc2db1987ad9e443ccdea91647abe2d567dc
15:57.48 INFO: Last bad revision: eb89ea3592d30a558f070889e88e2946ab9785a4
15:57.48 INFO: Pushlog:
https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=eb89ea3592d30a558f070889e88e2946ab9785a4&tochange=f8febc2db1987ad9e443ccdea91647abe2d567dc
f8febc2db1987ad9e443ccdea91647abe2d567dc Olli Pettay — Bug 1708042, use control priority for DidComposite but dispatch MozAfterPaint using mediumhigh, since scripts shouldn't run in control queue, r=bas
a0fccd7121b52f650cd27c93a0b4aa471fddbc9b Olli Pettay — Bug 1708042, add support for 'control' priority in ipdl, r=jld,ipc-reviewers
ddc6d95f0601deb688d0e0b4a6c186b8ddc013ab Olli Pettay — Bug 1708042, add control priority to the main thread, r=bas
Description
•