All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=917247

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=20312fc8acee901506323df29c7b6746884b03c802a43484494acf4fb9350437


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

mac-10_12_laptop_low_end-perf

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

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/149d6aee140000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/149d6aee140000

[PE] Avoid cache conflict when painting LocalFrameView outside of lifecycle by wangxianzhu@chromium.org
https://chromium.googlesource.com/chromium/src/+/81fc6554b2da741da30913dee5008cc83b20091d
memory:chrome:all_processes:reported_by_chrome:gpu:effective_size: 2.134e+07 → 2.603e+07 (+4.692e+06)

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

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

The fix for  bug 916860  didn't fix this regression.

 Issue 917236  has been merged into this issue.

 Issue 917246  has been merged into this issue.

I'm closing this bug because
1. The "private footprint" result is a progression (208.2MiB -> 196.2MiB)
2. The increased GPU memory is all free_size
3. The GPU memory increases for one web page only.