New issue
Advanced search Search tips
Note: Color blocks (like █ or █) mean that a user may not be available. Tooltip shows the reason.

Issue 733554 link

Starred by 2 users
Status: WontFix
Owner:
sprang@chromium.org
Closed: Jun 2017
Cc:
mbonadei@chromium.org
█ kjellander@chromium.org
terelius@chromium.org
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

1%-32.9% regression in webrtc_perf_tests at 18536:18538 due to Android NDK GCC->Clang switch

Project Member Reported by terelius@chromium.org, Jun 15 2017 
Encode time, CPU usage and dropped frames have increased. PSNR, SSIM and frame rate have decreased. The regression seems to be caused by a Chromium roll, but nothing stands out as particularly suspicious.

https://chromium.googlesource.com/external/webrtc/+/d3d8702bcf8cb263dbda758cd7c00e123cdba791

Comment 1 by terelius@chromium.org, Jun 15 2017 

All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=733554

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgtqWatgoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgtvCpuAsM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1ovO_QkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1s__6wkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1rb87wgM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDglvqLzwoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1vaK9QoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1veC4ggM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1t23qAsM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgtpbpqwoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1qmb9QkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDglq_pggoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1vmppAkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1uK0iQgM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1oGF6AoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDglpvN5woM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDglq_p_AgM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDglvPBmwkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1qzeogoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDglunE4AsM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDglr2fuAoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1szCoAkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDglr7A6ggM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1oj3tAkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1u6E7wgM


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

webrtc-android-tests-nexus4-lollipop
webrtc-android-tests-nexus5
webrtc-android-tests-nexus6-nougat
webrtc-android-tests-nexus72

Comment 2 by sprang@chromium.org, Jun 15 2017

Cc: mbonadei@chromium.org
Can't see anything suspicious in the blame list. The chromium roll doesn't contain any libvpx changes.

It is interesting that only the Android bots are affected though. Infra change?
+mbonadei@

Comment 3 by kjellander@chromium.org, Jun 15 2017

Status: WontFix (was: Assigned)
Summary: 1%-32.9% regression in webrtc_perf_tests at 18536:18538 due to Android NDK GCC->Clang switch (was: 1%-32.9% regression in webrtc_perf_tests at 18536:18538)
We haven't done any changes to the perf bots recently, so it's unlikely this is related to infrastructure. Especially as several devices are affected in a similar way.

Looking at the roll, it contains the Android GCC -> Clang switch (https://chromium.googlesource.com/chromium/src/build/+/5a7ba0fdfa1c5c98291d5a66f5d2a085051de215) which is known to affect performance (see  crbug.com/481675 ).

I think there's nothing we can or should do in this case except tracking it in a bug for future reference (+ a note in the sheriff diary)

Comment 4 by sprang@chromium.org, Jun 15 2017 

Ah, that'd explain it. I just saw the "No update to Clang." on the landing page and didn't dig further into the build diff. :)
►

Sign in to add a comment