All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=815154

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=f6493670df5648b99b7fb7144120f2b2ff54849a017659ae31b6961fd12ed5e7


Bot(s) for this bug's original alert(s):

chromium-rel-mac12

📍 Pinpoint job started.
https://chromeperf.appspot.com/job/12b2e120440000

😿 Pinpoint job stopped with an error.
https://chromeperf.appspot.com/job/12b2e120440000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/11b2a5ff840000

😿 Pinpoint job stopped with an error.
https://pinpoint-dot-chromeperf.appspot.com/job/11b2a5ff840000

Suspecting https://chromium.googlesource.com/chromium/src/+/28cc4cd37aa9c4e27d882c249092098197920133.

This result is strange. On my machine, that test case runs at 48fps in stable and 60fps in canary, due to my perf fix.
http://webglsamples.org/dynamic-cubemap/dynamic-cubemap.html
Wonder if there's something wrong with the metric or a pathological edge case on that test system.

This is probably GPU-Intel.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/23b2d93e0d72be12fe5366ff77fbcb41c5f95d96

commit 23b2d93e0d72be12fe5366ff77fbcb41c5f95d96
Author: Kai Ninomiya <kainino@chromium.org>
Date: Thu Mar 01 19:04:24 2018

Experimentally removing MSAA workaround for perf bot

The perf bot perf.chromium-rel-mac12 has alerted on a ~50% slowdown in
http___webglsamples.org_dynamic-cubemap_dynamic-cubemap.html .
In the regression range was my seemingly closely
related CL: https://crrev.com/c/923103
But that commit should improve performance, not regress it.

So this CL adds an exception to the workaround in 923103, so that
that workaround does not apply on perf.chromium-rel-mac12.
This should help determine whether my CL really caused a regression.

I'm expecting this CL to cause a regression in
http___kenrussell.github.io_webgl-animometer_Animometer_tests_3d_webgl.html
(undoing the recent improvement caused by 923103).

Bug:  815154 
Change-Id: I335133a127bb7cfcb92ba29065641dfd55b63166
Reviewed-on: https://chromium-review.googlesource.com/938844
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Zhenyao Mo <zmo@chromium.org>
Cr-Commit-Position: refs/heads/master@{#540237}
[modify] https://crrev.com/23b2d93e0d72be12fe5366ff77fbcb41c5f95d96/content/test/gpu/gpu_tests/pixel_expectations.py
[modify] https://crrev.com/23b2d93e0d72be12fe5366ff77fbcb41c5f95d96/gpu/config/gpu_driver_bug_list.json

Indeed, removing the workaround both fixes the regression on
http___webglsamples.org_dynamic-cubemap_dynamic-cubemap.html
and re-regresses
http___kenrussell.github.io_webgl-animometer_Animometer_tests_3d_webgl.html .

Since with max_msaa_sample_count_4:
* Animometer is better on the perf bot
* Animometer is better on my machine
* dynamic-cubemap is better on my machine
I'm going to revert my experiment.
(I also may need to revert it in the M66 branch if it went in there.)

I strongly suspect something is wrong with the dynamic-cubemap.html perf test,
or at least that it's hitting a very weird performance edge case on very particular hardware.

Adding labels in case I forget to remove this from M66 (if it made it into M66).

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/301e7863e403360983644f49506368a867a77728

commit 301e7863e403360983644f49506368a867a77728
Author: Kai Ninomiya <kainino@chromium.org>
Date: Fri Mar 02 18:42:53 2018

Revert "Experimentally removing MSAA workaround for perf bot"

This reverts commit 23b2d93e0d72be12fe5366ff77fbcb41c5f95d96.

Reason for revert: Experiment complete. I don't think this workaround should be kept.

Original change's description:
> Experimentally removing MSAA workaround for perf bot
>
> The perf bot perf.chromium-rel-mac12 has alerted on a ~50% slowdown in
> http___webglsamples.org_dynamic-cubemap_dynamic-cubemap.html .
> In the regression range was my seemingly closely
> related CL: https://crrev.com/c/923103
> But that commit should improve performance, not regress it.
>
> So this CL adds an exception to the workaround in 923103, so that
> that workaround does not apply on perf.chromium-rel-mac12.
> This should help determine whether my CL really caused a regression.
>
> I'm expecting this CL to cause a regression in
> http___kenrussell.github.io_webgl-animometer_Animometer_tests_3d_webgl.html
> (undoing the recent improvement caused by 923103).
>
> Bug:  815154 
> Change-Id: I335133a127bb7cfcb92ba29065641dfd55b63166
> Reviewed-on: https://chromium-review.googlesource.com/938844
> Commit-Queue: Kai Ninomiya <kainino@chromium.org>
> Reviewed-by: Kai Ninomiya <kainino@chromium.org>
> Reviewed-by: Zhenyao Mo <zmo@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#540237}

TBR=zmo@chromium.org,kainino@chromium.org

Change-Id: Iddc41a8cda141d2722a1fd9c1980604dfa61a0d3
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug:  815154 
Reviewed-on: https://chromium-review.googlesource.com/946608
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Cr-Commit-Position: refs/heads/master@{#540566}
[modify] https://crrev.com/301e7863e403360983644f49506368a867a77728/content/test/gpu/gpu_tests/pixel_expectations.py
[modify] https://crrev.com/301e7863e403360983644f49506368a867a77728/gpu/config/gpu_driver_bug_list.json

oops, didn't intend to land that No-Try. Keeping an eye on the builds to make sure removing those expectations didn't break something.

Requesting merge of the revert patch in #12. Merge CL opened here:
https://chromium-review.googlesource.com/c/chromium/src/+/947253

This bug requires manual review: Reverts referenced in bugdroid comments after merge request.
Please contact the milestone owner if you have questions.
Owners: cmasso@(Android), cmasso@(iOS), josafat@(ChromeOS), abdulsyed@(Desktop)

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

The NextAction date has arrived: 2018-03-05

Please add affected OSs.

Before we approve merge to M66, could you pls confirm followings?

Is the revert listed at #14 well baked/verified in Canary?
Any other imp details to justify the merge.

Both the version before the revert and after the revert are well tested.
This change is extremely specific to one piece of Mac hardware - all the original patch and the revert do is disable and re-enable a workaround entry for that hardware only.

The only reason there is a "revert" is because my temporary experiment (meant to affect our perf bots so I could check the performance result difference) was complete.

> Both the version before the revert and after the revert are well tested.

(baked/verified in Canary).

Approving merge to M66 branch 3359 based on comments #20, #21 and #22. Please merge ASAP so we can pick it up for tomorrow's M66 dev release. Thank you.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/0f69985237f0bf8ba146be39ba1b545cccff600f

commit 0f69985237f0bf8ba146be39ba1b545cccff600f
Author: Kai Ninomiya <kainino@chromium.org>
Date: Wed Mar 07 19:29:35 2018

Revert "Experimentally removing MSAA workaround for perf bot"

This reverts commit 23b2d93e0d72be12fe5366ff77fbcb41c5f95d96.

Reason for revert: Experiment complete. I don't think this workaround should be kept.

Original change's description:
> Experimentally removing MSAA workaround for perf bot
>
> The perf bot perf.chromium-rel-mac12 has alerted on a ~50% slowdown in
> http___webglsamples.org_dynamic-cubemap_dynamic-cubemap.html .
> In the regression range was my seemingly closely
> related CL: https://crrev.com/c/923103
> But that commit should improve performance, not regress it.
>
> So this CL adds an exception to the workaround in 923103, so that
> that workaround does not apply on perf.chromium-rel-mac12.
> This should help determine whether my CL really caused a regression.
>
> I'm expecting this CL to cause a regression in
> http___kenrussell.github.io_webgl-animometer_Animometer_tests_3d_webgl.html
> (undoing the recent improvement caused by 923103).
>
> Bug:  815154 
> Change-Id: I335133a127bb7cfcb92ba29065641dfd55b63166
> Reviewed-on: https://chromium-review.googlesource.com/938844
> Commit-Queue: Kai Ninomiya <kainino@chromium.org>
> Reviewed-by: Kai Ninomiya <kainino@chromium.org>
> Reviewed-by: Zhenyao Mo <zmo@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#540237}

TBR=kainino@chromium.org, zmo@chromium.org

Change-Id: Iddc41a8cda141d2722a1fd9c1980604dfa61a0d3
No-Try: true
Bug:  815154 
Reviewed-on: https://chromium-review.googlesource.com/946608
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Cr-Original-Commit-Position: refs/heads/master@{#540566}(cherry picked from commit 301e7863e403360983644f49506368a867a77728)
Reviewed-on: https://chromium-review.googlesource.com/947253
Cr-Commit-Position: refs/branch-heads/3359@{#69}
Cr-Branched-From: 66afc5e5d10127546cc4b98b9117aff588b5e66b-refs/heads/master@{#540276}
[modify] https://crrev.com/0f69985237f0bf8ba146be39ba1b545cccff600f/content/test/gpu/gpu_tests/pixel_expectations.py
[modify] https://crrev.com/0f69985237f0bf8ba146be39ba1b545cccff600f/gpu/config/gpu_driver_bug_list.json

Thanks, merged.

Marking the actual bug as WontFix as the outcome of my investigation was that this regression result is probably either not correct or an unusual edge case.

 Issue 880898  has been merged into this issue.

^ was an accidental merge into the wrong issue.