Issue metadata
Sign in to add a comment
|
3.3%-18% regression in webrtc_perf_tests at 13478:13479 |
||||||||||||||||||||
Issue descriptionSeveral audio performance values changed on a single bot without any obvious code change. Peak/blame CL is just an utility change, recent CLs look unrelated too. Can you take a deeper look at that bot?
,
Jul 20 2016
Yes, it would suggest 13479 or thereabouts is guilty. https://chromium.googlesource.com/external/webrtc/+/73ab917d27b6e66f075802c473cdec3d605ef5e5 doesn't seem very likely like you say. However, https://chromium.googlesource.com/external/webrtc/+/39607c9e3433389db0256d1df4fcdf58a99fedf8 (the CL before) changes mac code and the above regression is mac-only. That is very interesting. Also https://chromium.googlesource.com/external/webrtc/+/6c4d0f0bbf2522b7869f6238a08e27535a5adcb1 is pretty close. Danil, can you look up the owner of the test from blame info and ask them what the regressed numbers mean? Magnus, Henrik, cc:ing you since it's your CLs I mention above.
,
Jul 20 2016
,
Jul 20 2016
Per, this performance values are all about new Level Controller you added recently. Can you describe what their regression mean.
,
Jul 25 2016
My CL is platform agnostic and only adds periodic logs once every 10 seconds. Can't be related IMO.
,
Aug 5 2016
@peah, gentle ping. Any insights on this?
,
Aug 8 2016
I cannot see anything changing in the code that warrants this increase. Has there been any changes in the hardware running the tests?
,
Aug 15 2016
These are actually fine, since the tests do not yet do anything on Android. They will be resolved once the Android functionality is activated in the tests. |
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by danilchap@chromium.org
, Jul 19 2016