1421611 - https://codepen.io/thebabydino/full/LRZvdN/ performs badly and renders wrong

Status: RESOLVED WORKSFORME

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

Comment 1

[Darkspirit]

So many codepen testcases are not working anymore because their SCSS seems broken. I should try to inform them.

Comment 2

[Darkspirit]

Broken test case. Plead for INCOMPLETE.

Comment 3

[Markus Stange [:mstange]]

Looks like codepen's SCSS compiler has a bug. Here's a fork of the testcase that works: https://codepen.io/mstange/full/zWGYxq
(I compiled the SCSS using sassmeister.com and pasted the result as raw CSS.)

Comment 4

[Markus Stange [:mstange]]

Profile: https://perfht.ml/2FvhXQt

All the time is being spent in the content thread and in the render backend. The renderer thread is not very busy and is capable of sending a lot more frames to GL than it does.
All the animations on this page are OMTA animations. Once we have async scene processing for OMTA, the render backend's slowness will not affect the frame rate. And once we throttle the animation on the content side (bug 1419851), the render backend will not get distracted by unnecessary new display lists.

The render backend spends 45% of its time in memory allocation and deallocation.

Comment 5

[Hiroyuki Ikezoe (:hiro)]

So, bug 1419851 has been fixed, it might be worth taking a new profile on the latest nightly?

Comment 6

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

https://perfht.ml/2GCDJ4J is a new profile. But on my macOS nightly with WR enabled it seems to perform better than on beta with WR disabled. So maybe we can just close this as fixed.

Comment 7

[Hiroyuki Ikezoe (:hiro)]

That's good to know. :)

Comment 8

[Hiroyuki Ikezoe (:hiro)]

Forgot to leave more comments.

In the profile in comment 6, I noticed one thing we can optimize which is related to animations, I've filed bug 1463605 for that.

Closing.