Issue metadata
Sign in to add a comment
|
1%-48.2% regression in webrtc_perf_tests at 13626:13715 |
||||||||||||||||||||||
Issue descriptionSee the link to graphs below.
,
Aug 15 2016
mflodman@: I've done some digging on the causes of these perf regressions and I wonder if they could have been caused by the CL https://codereview.webrtc.org/2165743003 ? If not, do you have any other suggestion about what could have caused the regressions?
,
Aug 17 2016
I can't see that this CL should have any effect at all for this test, since the number of audio streams is 0 and this is an audio only change. Stefan, I'm assigning to you to decide the impact of this and next step for a possible investigation.
,
Aug 17 2016
Hmm, putting back on me for now to investigate.
,
Aug 17 2016
Per, Not sure I'm reading this correctly or not, but this is what I see: One of the two graphs points at something between 13625-13649, the CL above is 13527 and not in that range. The other is in the 13712-13715 range, so shouldn't be there either.
,
Aug 22 2016
holmer@: Could the webrtc-android-tests-nexus9 tests be related to the CL https://codereview.webrtc.org/2201093006? The webrtc-android-tests-nexus9 seems to possibly be related to that, but the webrtc-android-tests-nexus72 are likely caused by something else. Please reassign back to me after having a look at the webrtc-android-tests-nexus9 tests, and I'll then have another look into the webrtc-android-tests-nexus72 tests.
,
Aug 22 2016
It's probably related to that and therefore I think it's ok to close. I expected metrics like those to be affected. It's not impossible that the nexus7 metrics also are related to that CL.
,
Aug 22 2016
Great! Thanks! I'll then close this. |
|||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||
Comment 1 by peah@chromium.org
, Aug 13 2016