All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=814225

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=c0061248aed7207b3abb60cf659395fa58447cf5894d04bd58a3ed3c8c333ff6


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

android-webview-nexus5X
chromium-rel-mac11-air
chromium-rel-mac11-pro
chromium-rel-mac12

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/179f4217840000

📍 Found significant differences after each of 3 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/179f4217840000

Mac SelectFileDialog TLC by tapted@chromium.org
https://chromium.googlesource.com/chromium/src/+/4de65ba04a18745d4963ca386a3e3f2b1735135a

Revert "Mac SelectFileDialog TLC" by thomasanderson@chromium.org
https://chromium.googlesource.com/chromium/src/+/374de998ab969466b15844bca93580a2332056ef

[iOS] Disable fullscreen for SadTab. by kkhorimoto@chromium.org
https://chromium.googlesource.com/chromium/src/+/e8bbb37f4c9a26f147c5623ef88dd7828a53d83b

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

iOS does not use blink, so marking as available and removing iOS chromium contributors.

alexclarke: can we re-run this? That leaves a CL and its revert in the bisect. And it's a CL that can't affect Android, which also regressed. In fact these CLs are not even in the bisect given on the perf chart -- https://chromium.googlesource.com/chromium/src/+log/e748cd60525acb71a5109195c855d0bcd4dc05e8%5E..12a96fd3a734b8d6a12abf75cce313fe60f9f35c?pretty=fuller&n=1000

So something is weird here.

There's a canvas change, a skia roll, and an angle roll in that range - that's where I'd be looking.

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/15c77505440000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14f7f405440000

Lets see what the bots find.

📍 Found significant differences after each of 2 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/14f7f405440000

Mac SelectFileDialog TLC by tapted@chromium.org
https://chromium.googlesource.com/chromium/src/+/4de65ba04a18745d4963ca386a3e3f2b1735135a

Revert "Mac SelectFileDialog TLC" by thomasanderson@chromium.org
https://chromium.googlesource.com/chromium/src/+/374de998ab969466b15844bca93580a2332056ef

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

The pinpoint is failing because `gn gen` fails on mac between r537419 and r537449. can we run this on android-webview-nexus5X instead?

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/12a172ff440000

📍 Found significant differences after each of 2 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/12a172ff440000

cros: Highlight matched search query in KSV. by wutao@chromium.org
https://chromium.googlesource.com/chromium/src/+/e748cd60525acb71a5109195c855d0bcd4dc05e8

Roll src/third_party/angle/ 17448956b..0cbfa5866 (2 commits) by angle-chromium-autoroll@skia-buildbots.google.com.iam.gserviceaccount.com
https://chromium.googlesource.com/chromium/src/+/12a96fd3a734b8d6a12abf75cce313fe60f9f35c

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

😿 Pinpoint job stopped with an error.
https://pinpoint-dot-chromeperf.appspot.com/job/15c77505440000

Previous job on this bug failed due to https://github.com/catapult-project/catapult/issues/4393, retrying now that it's fixed.

📍 Pinpoint job started.
https://chromeperf.appspot.com/job/14a8c0dac40000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/15aa1519c40000

📍 Found significant differences after each of 2 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/15aa1519c40000

Mojofy spellcheck messages by xiaochengh@chromium.org
https://chromium.googlesource.com/chromium/src/+/d9085ee1dd4bcd9907ffc348679bd5bb2628e631

Roll AFDO from 66.0.3347.0_rc-r1 to 66.0.3350.0_rc-r1 by afdo-chromium-autoroll@skia-buildbots.google.com.iam.gserviceaccount.com
https://chromium.googlesource.com/chromium/src/+/abc55a5385dd933035f92bd71c2c0079f1b207f3

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/149790fbc40000

📍 Found significant differences after each of 2 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/149790fbc40000

Mojofy spellcheck messages by xiaochengh@chromium.org
https://chromium.googlesource.com/chromium/src/+/d9085ee1dd4bcd9907ffc348679bd5bb2628e631

Roll AFDO from 66.0.3347.0_rc-r1 to 66.0.3350.0_rc-r1 by afdo-chromium-autoroll@skia-buildbots.google.com.iam.gserviceaccount.com
https://chromium.googlesource.com/chromium/src/+/abc55a5385dd933035f92bd71c2c0079f1b207f3

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

Another AFDO roll pinpointed.

Thanks for the report!

AFDO doesn't apply to mac, so I'm kicking those off of this bug. Will look into the Android regression.

Yeah, okay, this appears to be two regressions in one.

The commit after the 20% regression ("Mojofy spellcheck messages") in https://pinpoint-dot-chromeperf.appspot.com/job/149790fbc40000 explicitly calls out that it only applies to Mac and Android, which fits nicely with the Mac complaints that I kicked off of this bug. (Oops)

The AFDO roll caused a regression, too, but more on the order of 5%, which is par for the course for AFDO rolls (we sample what Chrome does every day and tell the compiler how to optimized based on it. For macrobenchmarks, this is often relatively stable. For smaller benchmarks that're sensitive to optimization decisions like some of blink's, well...)

xiaochengh - is it expected for "Mojofy spellcheck messages" to cause something like this?

 Issue 843620  has been merged into this issue.