All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=792008

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=4050d6b3a8e1b15f0a084ba6be3604df944c2629eb29fa6b5b3c09d82fb3852b


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

android-nexus5
android-nexus7v2
chromium-rel-mac-retina
chromium-rel-mac11
chromium-rel-mac11-air

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8961039498926417856

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14c6e2d8840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/12c6e2d8840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/11f219a8840000

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/12c6e2d8840000

📍 Found significant differences after each of 3 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/11f219a8840000

Make sure to visit clips and dst proxies during gather
By csmartdalton@google.com · Thu Nov 30 12:14:02 2017 +0000
skia @ 8816b93134db2fecdfd690fd26967468f6c814b9

Revert "Revert "Fix rendering of drrects with small circular inner rrects.""
By bsalomon@google.com · Fri Dec 01 21:25:39 2017 +0000
skia @ 6574921253197338f879130ed99fb9ce483976de

Remove WTF::WeakPtr entirely and replace it with base::WeakPtr.
By slangley@chromium.org · Sat Dec 02 15:26:15 2017
chromium @ 85cb2f71a50dd7da960d31cd599df1cd9d6487b9

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

📍 Found significant differences after each of 14 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/14c6e2d8840000

Roll src/third_party/angle/ 76746f9bc..d5f44c986 (1 commit)
By angle-deps-roller@chromium.org · Sat Dec 02 12:06:35 2017
chromium @ 47f567e22ada9d1d0f67724e0b7a96c7ad753319

Make sure to visit clips and dst proxies during gather
By csmartdalton@google.com · Thu Nov 30 12:14:02 2017 +0000
skia @ 8816b93134db2fecdfd690fd26967468f6c814b9

Make cpu scaling more robust
By kjlubick@google.com · Thu Nov 30 14:13:22 2017 +0000
skia @ 1c8092ac64dc4d9fd140d7fd46fa52163fe26569

Add a GrContextThreadSafeProxy to SkSurfaceCharacterization
By robertphillips@google.com · Thu Nov 30 14:16:32 2017 +0000
skia @ 8def8bffe98ee58af6c088f3ab6e4a41f32b732f

Renaming and refactoring to prepare for init-once threaded backend
By liyuqian@google.com · Thu Nov 30 19:35:08 2017 +0000
skia @ 07a42411f8d4a00d992bafdd18f2183f58c4547a

Roll skia/third_party/externals/angle2/ 97fa85579..e218f15fc (1 commit)
By angle-deps-roller@chromium.org · Thu Nov 30 19:41:38 2017 +0000
skia @ 384295bac27f12b2a18141274be991a02f7b64a3

Add "-Wno-over-aligned" to Android.bp for x86 Android
By scroggo@google.com · Thu Nov 30 20:08:38 2017 +0000
skia @ 74663e722ea905972e62c59e60cd0cbc97e8435f

encode kAlpha_8 as grayalpha with sigbits for gray==1
By reed@google.com · Thu Nov 30 21:06:38 2017 +0000
skia @ d6cb11ee9df514a3a65b313725d3ac3bfd003e7f

Revert "Renaming and refactoring to prepare for init-once threaded backend"
By scroggo@google.com · Thu Nov 30 21:27:59 2017 +0000
skia @ 8a90c4da1cc7998927ce8f282b28baaf976a3873

remove read/write rawpixels
By reed@google.com · Thu Nov 30 21:43:08 2017 +0000
skia @ efe7c4920194796e997bff5bdb94e6fade6cb367

Fix rendering of drrects with small circular inner rrects.
By bsalomon@google.com · Fri Dec 01 20:00:09 2017 +0000
skia @ 096074af2c6fcbf622f0a42b2f7c56d0c66fbd80

Reland "Set multitexturing threshold for NVIDIA"
By bsalomon@google.com · Fri Dec 01 20:25:39 2017 +0000
skia @ bbf05759dedafcc924fe2247f2cbdd30688a3bfe

Initialize dst image memory for small sigma
By herb@google.com · Fri Dec 01 22:21:20 2017 +0000
skia @ 65f6985a96ef631b033f299fc6796122013782b8

Roll skia/third_party/externals/angle2/ 361df0703..76746f9bc (1 commit)
By angle-deps-roller@chromium.org · Fri Dec 01 23:49:19 2017 +0000
skia @ 0d05ca3e2fc36a8397b4cec4c23284beaf87642c

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

csmartdalton: Looking at the bisects, it's pretty clear the regression is due to your CL, "Make sure to visit clips and dst proxies during gather" (https://skia-review.googlesource.com/77940)

Can you take a look?

Not visiting clips and dst proxies was a bug. If there was a performance regression I'm pretty sure it was artificially high previously due to drawing incorrectly.

This area belongs more to robertphillips@ than me. Robert, thoughts? I think we should close this off as not a regression.

Right, this was to help fix skbug.com/7190 so I think we're going to have to live with it for now.

Long term we should regain some of this performance when Chrome is switched over to explicitly allocating resources and sorting render targets.