All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=808653

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=d3538faad47b835369783f528c202dae99a462cb20dfd23ad43f9e24428628d4


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

android-nexus6

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/12c316de840000

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/12c316de840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/16fd6121840000

😿 Pinpoint job stopped with an error.
https://pinpoint-dot-chromeperf.appspot.com/job/16fd6121840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14a554d9840000

😿 Pinpoint job stopped with an error.
https://pinpoint-dot-chromeperf.appspot.com/job/14a554d9840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/12a12675840000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/12a12675840000

viz: Tighten components/viz/service DEPS. by kylechar@chromium.org
https://chromium.googlesource.com/chromium/src/+/78ed387b857f368934b7c95757ac82afd9fbf329

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

I don't think changing DEPS include_rules made a performance difference. I hope not at least.

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/12f34fed840000

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/12f34fed840000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/11dde03d840000

This is a pretty big regression.

Marking available, and starting a bisect with a wider range. Looks to me like the first point after the regression was noisily low.

📍 Found significant differences after each of 2 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/11dde03d840000

[scheduler] Experiment to always treat input events as high-priority. by altimin@chromium.org
https://chromium.googlesource.com/chromium/src/+/ef42564665d915f2d2294d0bfa3a9ab5632c9777

Remaining changes to build all code in src.git with -Wimplicit-fallthrough. by thakis@chromium.org
https://chromium.googlesource.com/chromium/src/+/47d877575f265b60d1a412d6a01dc2d6eb8dc8d0

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

My change just added explicit fallthroughs and is perf and behavior preserving. If the bisect is correct, it has to be altimin's change. (But given the earlier comments on the bug, I'm not sure I trust this bisect)

Seems that something strange is happening with the metric.