All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=893790

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=e4bd754aeb11e36a502ee58676cd54144e73c197ea617737fc8deed9e5ef3fee


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

Android Nexus5X WebView Perf

blink_perf.canvas - Benchmark documentation link:
  https://bit.ly/blink-perf-benchmarks

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14b5b3d4e40000

📍 Found significant differences after each of 5 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/14b5b3d4e40000

On ES2 + EXT_color_buffer_half_float cannot render to GL_R16F. by bsalomon@google.com
https://skia.googlesource.com/skia/+/75975510f3bf5b1bfcf70a7f40695613bff0282e
toBlob_duration_jpeg: 1883 → 1909 (+25.35)

Add support for per-edge AA to GrTextureOp. by bsalomon@google.com
https://skia.googlesource.com/skia/+/2213ee94c7812cf06c5d7412ab7b9651f8587a58
toBlob_duration_jpeg: 1909 → 1740 (-168.8)

Remove obsolete paint copy by herb@google.com
https://skia.googlesource.com/skia/+/cb969c76e7331bf99ef4e5c6f3b064b503eeadb9
toBlob_duration_jpeg: 1740 → No values

Revert "Always include public/include headers for vulkan and just guard src files with SK_VULKAN." by mtklein@google.com
https://skia.googlesource.com/skia/+/684b506f3589ecc78eebe8d92082be31a5fbe348
toBlob_duration_jpeg: No values → 1685

Update canvas memory usage calculation by aaronhk@google.com
https://chromium.googlesource.com/chromium/src/+/4e1d4919de4b25e7a53edf095f3fdf2ddead0910
toBlob_duration_jpeg: 1680 → 2695 (+1015)

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

Benchmark documentation link:
  https://bit.ly/blink-perf-benchmarks

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/1282ad4ce40000

My change is to re-enable canvas memory tests and is not causing the regression. @fserb thinks that it is likely @bsalomon's change.

 Issue 893825  has been merged into this issue.

 Issue 893836  has been merged into this issue.

Actually, looking at the other regressions, I should take a closer look.

📍 Found significant differences after each of 4 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/1282ad4ce40000

Update canvas memory usage calculation by aaronhk@google.com
https://chromium.googlesource.com/chromium/src/+/4e1d4919de4b25e7a53edf095f3fdf2ddead0910
toBlob_duration_jpeg: 1694 → 2697 (+1003)

Fix undefined casts in sk_*_pop_free and sk_*_deep_copy. by davidben@google.com
https://boringssl.googlesource.com/boringssl/+/fb4e2e0f0c6547d29a53a48d7f6f131cbb936c1e
toBlob_duration_jpeg: 2706 → 2914 (+207.7)

Mostly fix undefined casts around STACK_OF's comparator. by davidben@google.com
https://boringssl.googlesource.com/boringssl/+/52483994c8f4e2861a071c20d0342970709d60a1
toBlob_duration_jpeg: 2914 → 2957 (+43.44)

Inline functions are apparently really complicated. by davidben@google.com
https://boringssl.googlesource.com/boringssl/+/a943613e4060fd19258e4ae82491af8f98d5e2a1
toBlob_duration_jpeg: 2964 → 2760 (-204)

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

Benchmark documentation link:
  https://bit.ly/blink-perf-benchmarks

(blink_perf tests tend not to run HTTPS at all, if I recall, so unlikely it's my changes.)

I'm unable to recreate any regression for `toBlob_duration_jpeg` on a Pixel 2 XL phone or my Linux desktop. Numbers are exactly the same with and without my CL, as it is only testing code for memory, that makes sense.

What do you think about this? We can't repro with our change.

I built the webview and shell apks locally on both ends of the full revision range from the original perf graph and can't reproduce a difference on toBlob_duration_jpeg.