Open Bug 1436316 Opened 8 years ago Updated 3 years ago

Negative translate serious performance regression

Categories

(Core :: Layout, defect, P3)

Product:
Core
Component:
Layout
Version:
58 Branch
Type:
defect

Tracking

()

Status:
UNCONFIRMED

People

(Reporter: p.borecki90, Unassigned)

Details

(Keywords: perf)

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/64.0.3282.140 Safari/537.36 Steps to reproduce: Navigate to http://idangero.us/swiper/demos/240-effect-coverflow.html using Firefox 58 (or 59) and resize window to reach width of about 1000px. Using mouse swipe the slides to the very end. Actual results: There is a regression from version 56. Performance drops significantly when negative translate3d on `.swiper-wrapper` exceeds windows width so first slides work as intended and last slides are stuttering heavily. e.g. for windows width of 1200px you can reach negative translate of -1200px before performance issues appear. Expected results: The swiper should perform the same for each slide. This is working as intended on firefox 56 and versions before that (not sure about 57).
Component: Untriaged → Layout
Keywords: perf
Product: Firefox → Core
This looks like we're throttling the animation incorrectly. Hiro, is this a known dupe?
Flags: needinfo?(hikezoe)
The animations have 'preserve-3d', so they have never run on the compositor thread (bug 779598), so if it's the reason the site should be slow on Firefox 56 too. And I don't see any shuttering there on my machine unfortunately. p.borecki90, could you please double check on currently nightly? If the shutter has gone, I guess it's thanks to bug 1425213.
Flags: needinfo?(hikezoe) → needinfo?(p.borecki90)
The bug is still there in the latest nightly (60.0a1 a.k.a. 60.0.0.6614), I'll today prepare a more appropriate demo on codepen/jsfiddle (the original demo tends to fail loading the images). I can even record a video showing the issue or upload performance recording if that will help with anything? Its very consistent, I'm getting it on three different machines so its not a local issue and on each one of them reverting to old version fixes it.
Flags: needinfo?(p.borecki90)
Thank you, p.borecki90. More reduced test case would help to know what's going on. If it could be reproducable without swipe, it would be really nice!
[ Triage 2017/02/20: P3 ]
Priority: -- → P3
I've had a rough two weeks so it took a `little` bit longer. Unfortunately I've encountered more of these regressions with the new engine (supposedly), I'll report them in separate issues. Reproduction link on jsfiddle: http://jsfiddle.net/zfegpfjx/62/
Sorry for the spam but I've also noticed two things, 1)Its not strictly negative translate related, same thing happends in positive values translate 2)Moving the mouse over animated elements somehow mitigates the issue forcing it to redraw with normal framerate
(In reply to p.borecki90 from comment #7) > Sorry for the spam but I've also noticed two things, > 1)Its not strictly negative translate related, same thing happends in > positive values translate > 2)Moving the mouse over animated elements somehow mitigates the issue > forcing it to redraw with normal framerate The second issue seems to be bug 1424506.
I've tested and its fully resolved in the latest 60.0b4 version (checked in jsfiddle example I've reported before), not sure within which bug report it was fixed but its now fixed nevertheless :)
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.