Prototype using PLS for clip task rendering on mobile devices with WR
Categories
(Core :: Graphics: WebRender, enhancement, P2)
Tracking
()
People
(Reporter: gw, Unassigned)
References
(Blocks 1 open bug)
Details
(Whiteboard: [wr-amvp][wr-q2])
Attachments
(1 file)
42.49 KB,
patch
|
Details | Diff | Splinter Review |
Using pixel local storage on supported devices seems like a perfect feature for clip mask rendering in WR. This has the potential to drastically reduce resolve bandwidth, and texture fetch bandwidth.
Reporter | ||
Updated•6 years ago
|
Updated•6 years ago
|
Updated•6 years ago
|
Reporter | ||
Comment 1•6 years ago
|
||
This patch contains a prototype of using pixel local storage for clip masks. It's incomplete, but has enough functionality to do benchmarking.
Unfortunately, it seems slower than the existing code on my test cases. This is surprising, since the memory resolve bandwidth should be much better.
I'm going to leave it here as a WIP patch for now, and revisit again in the near future.
It's possible that there are some mistakes in the patch, or some slow path that is being hit unexpectedly. It's also possible that in the test case I'm using the extra copy bandwidth is not relevant, but that it may be relevant in more real world cases.
Updated•6 years ago
|
Updated•6 years ago
|
Reporter | ||
Comment 2•5 years ago
|
||
Unassigning for now, since I'm not planning to work on this in the near future.
Updated•5 years ago
|
Updated•5 years ago
|
Reporter | ||
Updated•3 years ago
|
Description
•