[wayland] Enable Webrender on all compositors on Mesa Intel/AMD
Categories
(Core :: Graphics, enhancement)
Tracking
()
Tracking | Status | |
---|---|---|
firefox89 | --- | fixed |
People
(Reporter: rmader, Assigned: rmader)
References
(Blocks 1 open bug)
Details
Attachments
(1 file)
Following up on bug 1640115 comment 10
We do not yet enable Wayland by default anywhere and we only need to care for four major compositors: Weston, Mutter, KWin and Sway/wlroots.
Weston is only for testing purposes, Mutter has been well tested on Fedora for a while now, Sway appears to be fine (bug 1640115 comment 10) and KWin only enters a wider audience right now.
Lets enable Webrender on all of them and not invest more into the basic backend.
Open question: should we fall back to SW-WR instead of basic already if HW-WR fails?
Assignee | ||
Comment 1•4 years ago
|
||
Martin, Vlad, Andrew - any resistance again this?
Comment 3•4 years ago
•
|
||
Open question: should we fall back to SW-WR instead of basic already if HW-WR fails?
In an impromptu test, the total CPU usage for playing a YouTube video seems to increase from ~100% combined (gfx.webrender.force-disabled
) to ~160% combined (gfx.webrender.software
) here.
PS: I wonder if this is due to duplication of work in both Renderer and SWCompositor.
Fedora uses WebRender with KWin/Wayland already so no objections here. IMHO wayland compositor type is not as important as driver/mesa versions.
Assignee | ||
Comment 5•4 years ago
|
||
(In reply to Jan Alexander Steffens [:heftig] from comment #3)
Open question: should we fall back to SW-WR instead of basic already if HW-WR fails?
In an impromptu test, the total CPU usage for playing a YouTube video seems to increase from ~100% combined (
gfx.webrender.force-disabled
) to ~160% combined (gfx.webrender.software
) here.PS: I wonder if this is due to duplication of work in both Renderer and SWCompositor.
Getting SW-WR to perform as good as basic is AFAIK not a design goal per-se. IIUC one major motivation is that having only WR unlocks a bunch of possible optimizations (e.g. dropping display lists and much of the associated costs) - see for example bug 1701437 comment 2.
So slightly less performance in SW-WR should probably not be a blocker, as long as it's not too bad and the stability doesn't suffer.
Assignee | ||
Comment 6•4 years ago
|
||
This effectively enables WR on Weston, Sway and KWin-Wayland -
unlike on X11, few issues appear to be related to WR.
Updated•4 years ago
|
Comment 7•4 years ago
|
||
+1. For what it's worth, in Plasma/Wayland, Firefox works better with WR than with BR.
Assignee | ||
Updated•4 years ago
|
Comment 10•4 years ago
|
||
bugherder |
Updated•3 years ago
|
Description
•