All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=826820

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=a8f9a5a64280921165e05a8a04c01f46abb6c83e0e659094d201830efe08a8ec


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

android-nexus5
android-nexus6
android-one

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/148a93a7440000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/148a93a7440000

Trigger GPU context dropping on low-end from Browser proc by ericrk@chromium.org
https://chromium.googlesource.com/chromium/src/+/73cfd3fcb696344b7ecc66519c3a702f0ef6e127

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

My CL re-enabled an optimization that reduces GPU memory in the background. It brings us back to our previous levels before this optimization was temporarily disabled last month. Note that the "regression" here is ~500k on a few sites (on most sites we don't see it), and it comes with a large (multiple MB) reduction in system GPU memory, so it's an overall win:
https://chromeperf.appspot.com/report?sid=0246b0eeee7826261d963f2dfe3aab10a308845f010fb6da2cb4f24d8241c64d&start_rev=1519234338&end_rev=1522360965

We should understand why we have this additional usage at all, so not closing this out, but removing the regression / sheriff tags.

It appears that we are creating an additional command buffer that we didn't previously have - maybe this is in response to dropping contexts?

Issue 826832 has been merged into this issue.