All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=880384

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=9ae4d46b436c76949f4b670b9cdc092a23eac23f8f3057ad8e25f58589ea56b7


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

mac-10_12_laptop_low_end-perf

memory.desktop - Benchmark documentation link:
  None

system_health.memory_desktop - Benchmark documentation link:
  https://bit.ly/system-health-benchmarks

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/144ec713640000

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/144ec713640000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/12bbd82f640000

📍 Found significant differences after each of 4 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/12bbd82f640000

Cut-and-paste typo was limiting concurrency by allanmac@google.com
https://skia.googlesource.com/skia/+/4f28d76c010ee0f441beaffce223c8521deb9674
1.567e+07 → No values

Add back tight scaler context managment by herb@google.com
https://skia.googlesource.com/skia/+/ee2a3f36dcf7fbd4db664f93a08c095a8a76b20b
No values → 1.57e+07

Fix drawRect by robertphillips@google.com
https://skia.googlesource.com/skia/+/8c8b0462df8b3548d0d2b6c4287134133474e4b4
1.566e+07 → 1.54e+07 (-2.585e+05)

Mac OOP-D: Enable OOP-D for Mac Waterfall by fsamuel@chromium.org
https://chromium.googlesource.com/chromium/src/+/00ec720decf72b496b1b5bd5bd16bcc85d04ec03
1.6e+07 → 2.246e+07 (+6.462e+06)

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

Benchmark documentation link:
  https://bit.ly/system-health-benchmarks

Passing along to robertphillips@

Doesn't Robert's change reduce memory (-2.585e+05) use? BTW Robert is out until next year.

herb@ memory:chrome:all_processes:reported_by_chrome:skia:effective_size_avg / browse_media /

The regressions seem skia related. I don't think they're caused by my CL.

Pinpoint graph: https://pinpoint-dot-chromeperf.appspot.com/job/12bbd82f640000

Pinpoint shows that my cl increased memory use by 0.2%. It also shows that your change:

Commit 586455
Mac OOP-D: Enable OOP-D for Mac Waterfall

memory use increased from about 16M to 22M.

And since this is a Mac test I think pinpoint's analysis make sense.

Assigned to kylechar@ for triage.

Turning on OOP-D caused a ton of noise in memory tracing due to changing what process memory was allocated in, lots of graphics timing changes and the fact that it broke some memory reporting code. In particular there were problems with association of gpu and skia MemoryAllocatorDumps that could have impacted this. I investigating similar regressions and fixed broken memory reporting code in  crbug.com/876508 .

The regression caught by pinpoint has more than recovered and it's been four months so I don't think there is much to do here.