All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=734546

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg9pHe8woM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgtreV-woM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg9p-TqwoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgtpPM2QkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1vW6kgkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgttWtlAkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgtviLiwkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgtuDD5wkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgtsa8rAgM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg9uu8vQoM


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

webrtc-android-tests-nexus5
webrtc-android-tests-nexus5-kitkat
webrtc-android-tests-nexus72
webrtc-mac-large-tests
webrtc-win-large-tests

Ilya, these are the expected improvements?

There are many more alerts which should be linked to this issue, but at the moment I'm having some difficulty getting along with the tools.

Looks as expected, psnr and ssim should jump because there was a bug in how they were calculated before (dropped frames were compared to black frames). Also, because there are less comparisons, cpu usage should go down a little, this may improve decode and encode time on weak machines.

All changes make sense.