New issue
Advanced search Search tips

Issue 894450 link

Starred by 1 user
Status: Fixed
Owner:
sprang@chromium.org
Closed: Oct 11
Cc:
ilnik@chromium.org
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

1.1%-28% regression in webrtc_perf_tests at 25088:25092

Project Member Reported by ilnik@chromium.org, Oct 11 
This CL decreased PSNR and number of dropped frames. Also small change to media bitrate and decode time. All changes for screenshare only:

https://webrtc.googlesource.com/src/+/dcc023816eed052349578d41321d0a3cfdeb518a

Erik, I don't see why this should happen. We've probably broken some implicit assumption for the libvpx encoder by that change.
Project Member

Comment 1 by 42576172...@developer.gserviceaccount.com, Oct 11 

All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=894450

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=f7472ac9d1360ba8abc3aed1a088a02e95c8c74e65265f05627111c24d53bf80


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

webrtc-win-large-tests

Comment 2 by sprang@chromium.org, Oct 11

Status: Fixed (was: Assigned)
I think this is due to libvpx rate controler previously thinking it has 1s/fps extra buffer space when the re-encode is done.
psnr/ssim is down for sure, due to media bitrate being down. Number of dropper frames is however also down as are most delay metrics.
I'll take this as wai.
►

Sign in to add a comment