Issue metadata
Sign in to add a comment
|
2D canvas slow down with hardware acceleration
Reported by
palomath...@gmail.com,
Mar 19 2018
|
||||||||||||||||||||||||
Issue descriptionUserAgent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_11_6) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/65.0.3325.162 Safari/537.36 Steps to reproduce the problem: 1. High-resolution 2D canvas drawing using requestAnimationFrame() with hardware acceleration on a macbook pro with retina display What is the expected behavior? Performance around 60FPS What went wrong? Performance used to be around 60FPS but has dramatically dropped. Turning off hardware acceleration or reducing the resolution improves performance. Works as expected in FF and Safari without modification. Did this work before? Yes Previous version Chrome version: 65.0.3325.162 Channel: stable OS Version: OS X 10.11.6 Flash Version: Other people have encountered similar issues: https://productforums.google.com/forum/#!topic/chrome/4JklF86fFLk
,
Mar 20 2018
,
Mar 20 2018
palomathedoveborn@ Thanks for the issue. Tested this issue on MacBook Pro OS 10.13.1 with Retina display on the reported version 65.0.3325.162 and the latest Canary 67.0.3376.1 by following the below steps. 1. Launched Chrome with Hardware Acceleration Enabled and navigated to the link given in the original comment. 2. Navigated to the link http://poe.rivsoft.net/ and enabled FPS meter in Devtools -> Console -> Rendering. 3. On hovering the mouse, can observe the FPS ~20. The same behavior is observed when the Hardware Acceleration is Disabled. Attached is the screen cast for request. Request you to check and confirm if these are the right steps in reproducing the issue, which will help in further triaging. Thanks...
,
Mar 20 2018
I have had people try to replicate it but with no success. I think it may be an issue with my OSX being old (10.11.6). I attach a video of what I experience with the canvas example at https://al-ro.github.io/projects/heart/
,
Mar 20 2018
Thank you for providing more feedback. Adding the requester to the cc list. For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
,
Mar 21 2018
palomathedoveborn@ Thanks for the feedback. As per comment #4, the issue can be reproduced on Mac OS 10.11.6. As Mac OS 10.11.6 is unavailable at TE end to test this issue, adding 'TE-NeedsTriageFromMTV' label and requesting someone from MTV team to look into this issue and help in further triaging. Thanks..
,
Mar 22 2018
We've just seen this same issue with our Canvas app and after some troubleshooting believe it may have to do with El Capitan (OSX 10.11.6) and the latest Chrome version (65.0.3325.181) not playing well together. Echoing Comment 4. With High Sierra (OSX 10.13.3) and the same Chrome version (65.0.3325.181), canvas/HTML5 rendering with hardware acceleration ON render performance does not appear to be impacted, or at least not as severely. Overall it is a pretty crippling issue and hopefully the Chrome folks can take a look ASAP. Thanks!
,
Mar 23 2018
,
Mar 23 2018
Issue 824725 has been merged into this issue.
,
Apr 1 2018
Issue 822417 is potentially a duplicate? https://bugs.chromium.org/p/chromium/issues/detail?id=822417
,
Apr 1 2018
For the record, I'm also on El Capitan (OSX 10.11.6) and am seeing the same sluggish canvas performance.
,
Apr 5 2018
Also affecting users of https://www.piskelapp.com (a web based sprite editor) as confirmed on https://github.com/piskelapp/piskel/issues/796 Apparently makes the website completely unusable for users with OSX 10.11.6 (edit: reposted comment to fix incorrect link)
,
Apr 9 2018
This is also affecting users of http://p5js.org/ as seen in this issue: https://github.com/processing/p5.js/issues/2769
,
Apr 11 2018
Is it related with https://bugs.chromium.org/p/chromium/issues/detail?id=794819?
,
Apr 16 2018
vamshi.kommuri@ Please image one of the test machines with Mac 10.11 and retriage.
,
Apr 17 2018
cc'ing folks on related bug. Seems plausible that very broad GPU blacklisting caused a large perf regression. Can the change in skia that exposed the driver bug be dealt with instead?
,
Apr 17 2018
M67 Beta promotion is coming VERY soon. Pls make sure to land the fix and request a merge into the release branch ASAP. Thank you.
,
Apr 17 2018
ccameron@: Could you take a look? Any ideas?
,
Apr 17 2018
This issue was originally reported on M65, so exists on M66 and M67. Don't think this should be a beta blocker for M67. Pls let me know if I'm missing anything here.
,
Apr 17 2018
Please post the contents of about:gpu from an affected machine. It seems likely this is the same problem as Issue 773705 , where GPU acceleration for 2D canvas had to be disabled because of bugs in Apple's graphics driver for NVIDIA GPUs in 10.13 (apparently fixed in 10.13.4, and re-enabled in Chrome from that release on).
,
Apr 17 2018
I bet that this is a duplicate of issue 791828 . We blacklisted LOTS of GPUs gpu-raster and gpu-canvas in 65, and there are performance potholes in the non-accelerated path.
,
Apr 17 2018
This is also being reported on 10.11, and I'm not re-enabling GPU-anything there (just 10.13.4). But hopefully Kushal's patch can be merged to 67 and will fix this (testing it locally now).
,
Apr 18 2018
M67 Beta promotion is coming VERY soon. Your bug is labelled as Beta ReleaseBlock, pls make sure to land the fix and request a merge into the release branch latest by 1:00 PM PT Friday, 04/20. Thank you.
,
Apr 18 2018
Fix should land tonight, hopefully.
,
Apr 18 2018
Ok, great. Thank you. Pls request a merge to M67 once change is baked in canary.
,
Apr 19 2018
Just checking back, is the change landed in trunk last night?
,
Apr 19 2018
Duplicating to issue 773705 , which is the root cause (fix in the CQ).
,
Apr 19 2018
|
|||||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||||
Comment 1 by palomath...@gmail.com
, Mar 19 2018