New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 691741 link

Starred by 2 users

Issue metadata

Status: Verified
Owner:
Last visit > 30 days ago
Closed: Feb 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 1
Type: Bug-Regression

Blocking:
issue 612542
issue 687180



Sign in to add a comment

Webgl aquarium perf regression

Project Member Reported by marc...@chromium.org, Feb 13 2017

Issue description

https://chromeperf.appspot.com/report?sid=ae1b000d21e7b4f5205a70747a186cea75128ebd9703939d1e49173833076a1d&rev=30070000926700000

It seems like webgl aquarium performance regressed on the following intel devices:

- kefka, celes [braswell on kernel 3.18]
- samus, auron_yuna, auron_pain [broadwell on kernel 3.14]

 
Tested on 57.0.2941.0 (platform 9052) and the performance is fine.

Tested on 58.0.3006.0 (platform 9265) and the performance is fine.
Tested on 58.0.3006.0 (platform 9266) and the performance is fine.

Tested on 58.0.3007.0 (platform 9267) and the performance regresses

Comment 5 by kbr@chromium.org, Feb 14 2017

Cc: zmo@chromium.org qiankun....@intel.com
Components: Blink>WebGL
Labels: OS-Chrome
Qiankun: FYI, https://codereview.chromium.org/2680703003 is the current likely suspect. Unfortunately, we'll have to revert and ask you to re-land if that's the one.

Comment 6 by kbr@chromium.org, Feb 14 2017

Labels: -Type-Bug Type-Bug-Regression
Yup that's the root cause. https://codereview.chromium.org/2695833003/
Owner: marc...@chromium.org
Status: Started (was: Untriaged)

Comment 9 by kbr@chromium.org, Feb 14 2017

Components: Internals>GPU>Internals
Labels: GPU-Intel
Project Member

Comment 10 by bugdroid1@chromium.org, Feb 14 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/2404c8ae0098fa7cae48db92222240293a9d2689

commit 2404c8ae0098fa7cae48db92222240293a9d2689
Author: marcheu <marcheu@chromium.org>
Date: Tue Feb 14 02:48:52 2017

Revert of Select correct copy method for DoCopySubTexture (patchset #1 id:1 of https://codereview.chromium.org/2680703003/ )

Reason for revert:
The new DoCopySubTexture causes a performance regression on WebGL on Chrome OS.

BUG= chromium:691741 
TEST=on celes, webgl performance is back to normal

Original issue's description:
> Select correct copy method for DoCopySubTexture
>
> BUG= 687180 
> CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
>
> Review-Url: https://codereview.chromium.org/2680703003
> Cr-Commit-Position: refs/heads/master@{#448905}
> Committed: https://chromium.googlesource.com/chromium/src/+/a8450b810e89819add240a319586c899683dff9f

TBR=kbr@chromium.org,zmo@chromium.org,yunchao.he@intel.com,dongseong.hwang@intel.com,qiankun.miao@intel.com
# Not skipping CQ checks because original CL landed more than 1 days ago.
BUG= 687180 

Review-Url: https://codereview.chromium.org/2695833003
Cr-Commit-Position: refs/heads/master@{#450212}

[modify] https://crrev.com/2404c8ae0098fa7cae48db92222240293a9d2689/gpu/command_buffer/common/gles2_cmd_utils.cc
[modify] https://crrev.com/2404c8ae0098fa7cae48db92222240293a9d2689/gpu/command_buffer/common/gles2_cmd_utils.h
[modify] https://crrev.com/2404c8ae0098fa7cae48db92222240293a9d2689/gpu/command_buffer/service/gles2_cmd_apply_framebuffer_attachment_cmaa_intel.cc
[modify] https://crrev.com/2404c8ae0098fa7cae48db92222240293a9d2689/gpu/command_buffer/service/gles2_cmd_copy_texture_chromium.cc
[modify] https://crrev.com/2404c8ae0098fa7cae48db92222240293a9d2689/gpu/command_buffer/service/gles2_cmd_decoder.cc

Status: Fixed (was: Started)

Comment 12 by kbr@chromium.org, Feb 14 2017

Blocking: 687180 612542
How can I check if the performance was back with above CL?

Comment 15 by ihf@chromium.org, Feb 17 2017

Chrome hasn't upreved yet, otherwise it should be visible on the first links graph.

Comment 16 by ihf@chromium.org, Feb 18 2017

Chrome 3015 is good which contains your change in #13
https://chromium.googlesource.com/chromium/src/+log/58.0.3011.0..58.0.3015.0?pretty=fuller&n=10000

Comment 17 by dchan@google.com, Apr 17 2017

Labels: VerifyIn-59
Status: Verified (was: Fixed)
Verified on 9460.50.0/59.0.3071.71

Sign in to add a comment