Issue metadata
Sign in to add a comment
|
24%-31.1% regression in webrtc_perf_tests at 15441:15511 |
||||||||||||||||||||
Issue descriptionSee the link to graphs below.
,
Dec 13 2016
Hey ivoc, Would you mind having a look at this alert? I can't find a reasonable culprit CL in the blamelist, and I think you know more about this test than I do :) Note that all of the alerts are on the Nexus9 bot, so it may very well be a flakey result. (This should not be one of the bots with new devices, as per kjellander's email to the webrtc-perf-sheriffs email list.)
,
Dec 13 2016
Hmm, that is very strange, the code that's used in this test did not change at all, so I don't see how this can possibly be caused by any changes to our code. I think there is a problem with these tests that causes them to be unreliable (Per had similar problems with the level estimator perf tests). I will try to make some changes to the way these tests work to improve their reliability.
,
Dec 14 2016
Thanks for looking into it! I have added some more alerts that came in today to the group report.
,
Dec 14 2016
The following revision refers to this bug: https://chromium.googlesource.com/external/webrtc.git/+/930959d261451af05c31db944ecc440808eb89f8 commit 930959d261451af05c31db944ecc440808eb89f8 Author: ivoc <ivoc@webrtc.org> Date: Wed Dec 14 13:48:16 2016 Improvements to the reliability of the echo detector perf test. BUG= chromium:673683 Review-Url: https://codereview.webrtc.org/2568883004 Cr-Commit-Position: refs/heads/master@{#15606} [modify] https://crrev.com/930959d261451af05c31db944ecc440808eb89f8/webrtc/modules/audio_processing/residual_echo_detector_complexity_unittest.cc
,
Dec 14 2016
I hope this change will make the perf stats more robust, however, since this modifies the behavior of the stats it's expected to see a lot of alerts popping up.
,
Dec 15 2016
|
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by brandtr@chromium.org
, Dec 13 2016