All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=776172

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=06360ad8425bbfc1851ac38e9ffb311cd035cf49d3ef1ee028ea0b4ecab82564


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

android-nexus6

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8965355615746327136

=== Auto-CCing suspected CL author steimel@chromium.org ===

Hi steimel@chromium.org, the bisect results pointed to your CL, please take a look at the
results.


=== BISECT JOB RESULTS ===
Perf regression found with culprit

Suspected Commit
  Author : Tommy Steimel
  Commit : 07624e078141bbbe3612245d0ae57c617601c15f
  Date   : Tue Oct 17 04:06:09 2017
  Subject: Show sound content setting in Page Info when tab is audible on Android

Bisect Details
  Configuration: android_nexus6_perf_bisect
  Benchmark    : media.mobile
  Metric       : memory:chrome:all_processes:reported_by_os:system_memory:java_heap:proportional_resident_size_avg/video.html?src_tulip2.vp9.webm_background
  Change       : 2.05% | 16179712.0 -> 16511146.6667

Revision             Result                   N
chromium@509275      16179712 +- 87750.9      6      good
chromium@509277      16226475 +- 88981.0      6      good
chromium@509278      16608939 +- 116877       6      bad       <--
chromium@509279      16557739 +- 89676.5      6      bad
chromium@509282      16532480 +- 88751.9      6      bad
chromium@509289      16568661 +- 39144.8      6      bad
chromium@509303      16537088 +- 75661.8      6      bad
chromium@509330      16511147 +- 124477       6      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=video.html.src.tulip2.vp9.webm.background media.mobile

More information on addressing performance regressions:
  http://g.co/ChromePerformanceRegressions

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8965355615746327136


For feedback, file a bug with component Speed>Bisection

+miu@, +dalecurtis@, +mlamouri@. Looks like enabling the audio monitoring on Android has had a measurable effect on performance. Do you think this is enough of a problem to warrant re-disabling audio monitoring on Android?

Hmm, may have had a power impact, but the results are too noisy to tell, https://chromeperf.appspot.com/report?sid=7197cb5a6c6a5aa929fbbedec38a36df5d299b126e67821da9a6fe1ff1e39b80&start_rev=507308&end_rev=509877

I wouldn't recommend enabling it unless it's actually going to be used / visible regularly. 

FWIW, it was enabled to be used for  crbug.com/774256 

Hmmm...the regression was 2% memory use...that's rather strange. This could be pointing at an issue that affects desktop as well. We should understand why memory usage regressed so much, since the audio scanning doesn't actually use more memory.


As for power/CPU usage, perhaps there is a way to turn it on only while the UI is showing? This may require some extensive plumbing, however.

-or-

Perhaps the cost is so great because our ARM intrinsics implementation needs to be better optimized?

 Issue 776754  has been merged into this issue.

 Issue 776924  has been merged into this issue.

Spoke with Mounir and he thinks the  crbug.com/774256  feature on Android may not be worth a performance regression, so I think I'll just revert it unless @dalecurtis you think this is something your team would be able to look into and fix?

Sounds that, based on comment #5, dalecurtis@ might agree with the decision on comment #10 but assigning to them to confirm.

I'm not sure why more memory might be used. Nothing we do in that code should take much memory; certainly not ~500k more. I can only assume this is paging in previously dead code or causing some ancillary effects in the WebContents.

On performance I can only offer the suggestion I did last time, which is that we don't need the full EWMA algorithm for this work. Just a value != 0 test should be fine and is much cheaper.

If you can reproduce these issues locally I recommend shunting off the tab invalidation notifications and trying the v!=0 test to see if they have a positive effect.

But yes otherwise if this isn't going to provide a benefit to most users any regression here seems unwise.

+primiano@ in case of they can help with their memory skillz.

I think the memory regression could be a side effect of delaying the GC because of the audio level monitoring activity. You don't reach idleness and that might push the system to push the GC further down the road.

I'd be quite worried about the cpu usage though, the regression there is 12% on cpu time ^__^

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/0c9e59a7d6b7f5b307791efdbb98f512f03ed750

commit 0c9e59a7d6b7f5b307791efdbb98f512f03ed750
Author: Tommy Steimel <steimel@chromium.org>
Date: Fri Oct 27 18:11:08 2017

Revert "Show sound content setting in Page Info when tab is audible on Android"

This reverts commit 07624e078141bbbe3612245d0ae57c617601c15f.

Reason for revert: Enabling audio monitoring has had significant performance impact on Android. See  crbug.com/776172 

Original change's description:
> Show sound content setting in Page Info when tab is audible on Android
> 
> This CL updates Android to show the sound content setting in the Page
> Info Bubble when sound is playing or has recently played on a site. In
> order to use WebContents::WasRecentlyAudible() for this, we also enable
> monitoring audio levels on Android.
> 
> Bug:  774256 
> Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
> Change-Id: Ia00909f6a6343f1e2fabf2ee3d4e651b3640c81e
> Reviewed-on: https://chromium-review.googlesource.com/719707
> Reviewed-by: Mounir Lamouri <mlamouri@chromium.org>
> Reviewed-by: Ted Choc <tedchoc@chromium.org>
> Reviewed-by: Yuri Wiitala <miu@chromium.org>
> Commit-Queue: Tommy Steimel <steimel@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#509278}

TBR=miu@chromium.org,mlamouri@chromium.org,tedchoc@chromium.org,steimel@chromium.org

# Not skipping CQ checks because original CL landed > 1 day ago.

Bug:  774256 ,  776172 
Change-Id: Ifb33a6b779826d3416adb82cdd80ad56589eba68
Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
Reviewed-on: https://chromium-review.googlesource.com/739883
Reviewed-by: Ted Choc <tedchoc@chromium.org>
Reviewed-by: Mounir Lamouri <mlamouri@chromium.org>
Commit-Queue: Tommy Steimel <steimel@chromium.org>
Cr-Commit-Position: refs/heads/master@{#512238}
[modify] https://crrev.com/0c9e59a7d6b7f5b307791efdbb98f512f03ed750/chrome/browser/ui/android/page_info/page_info_popup_android.cc
[modify] https://crrev.com/0c9e59a7d6b7f5b307791efdbb98f512f03ed750/chrome/browser/ui/android/page_info/page_info_popup_android.h
[modify] https://crrev.com/0c9e59a7d6b7f5b307791efdbb98f512f03ed750/media/audio/audio_output_controller.h