All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=621169

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICggo2KvgoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICggun-uwsM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICggovT-wgM


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

chromium-rel-win7-gpu-ati

Kicked off bisect but there is only one media related commit in the range [1] that seems remotely related:
"Support HEVC through EME" https://codereview.chromium.org/1909313002

/cc servolk@


[1] 399930 - 399973 http://test-results.appspot.com/revision_range?start=399930&end=399973)

HEVC (and ENABLE_HEVC_DEMUXING buildflag) is only enabled on Chromecast platform. So I don't think my CL might affect anything, given that most changes there were within #if ENABLE_HEVC_DEMUXING sections.

Moving this nonessential bug to the next milestone.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Kicked off another bisect here:
https://chromeperf.appspot.com/buildbucket_job_status/9007410079705960544

I'd rather close this bug because results looks very flaky. It may be affected by surrounding environments, and this seems to be a kind of tests that need reference running.

After the regression, actually data was improved and regressed sometimes, and recent values looks better than values this regression happened.