All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=724053

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgmpqD6AkM


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

android-webview-nexus5X

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8979256144462316320

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

Hi binji@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 : binji
  Commit : e94f5343100836bcdc838dc4da4fd13d4c69a950
  Date   : Fri May 12 22:16:05 2017
  Subject: Enable SharedArrayBuffer feature by default.

Bisect Details
  Configuration: android_webview_arm64_aosp_perf_bisect
  Benchmark    : system_health.memory_mobile
  Metric       : memory:webview:all_processes:reported_by_chrome:v8:effective_size_avg/background_social/background_social_facebook
  Change       : 0.77% | 3045712.0 -> 3069176.0

Revision             Result                  N
chromium@471417      3045712 +- 0.0          6      good
chromium@471452      3045712 +- 0.0          6      good
chromium@471457      3045640 +- 394.36       6      good
chromium@471459      3045712 +- 0.0          6      good
chromium@471460      3069320 +- 0.0          6      bad       <--
chromium@471461      3069248 +- 394.36       6      bad
chromium@471470      3069248 +- 394.36       6      bad
chromium@471487      3069176 +- 498.831      6      bad

Please refer to the following doc on diagnosing memory regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/memory-infra/memory_benchmarks.md

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-webview --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=background.social.facebook system_health.memory_mobile

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8979256144462316320

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5343376653680640


| O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq
|  X  | for more information addressing perf regression bugs. For feedback,
| / \ | file a bug with component Speed>Bisection.  Thank you!

 Issue 725009  has been merged into this issue.

Ben, can you take a look at this?
Any idea why this would impact android memory?

This looks likely to be the additional builtins/codestubs for Atomics and SharedArrayBuffer per native context. I tried to find any other reason for the size increase, but this seems to be it. Unfortunately, I'm not sure there's much else that can be done here.

Given the above, I would venture that we should mark this as WontFix, bradnelson@, danno@ WDYT?

Thanks to some suggestions from adamk, I was able to find that the heap size is larger than it should be. It won't completely remove the size increase (since that is just what happens we add new builtins) but it should make it a little better.

Gonna work on that CL today.

The following revision refers to this bug:
  https://chromium.googlesource.com/v8/v8.git/+/284a4804f2f2cab172d2b4d96048b5d8986c72cd

commit 284a4804f2f2cab172d2b4d96048b5d8986c72cd
Author: Ben Smith <binji@chromium.org>
Date: Tue Jun 13 23:45:44 2017

[SAB] Move creation of SharedArrayBuffer/Atomics to InitializeGlobal

It is only attached to the global object if the --harmony-sharedarraybuffer
flag is enabled, but this allows more objects to be added to the snapshot which
seems to reduce the amount of heap memory used per context.

Bug:  chromium:724053 
Change-Id: I5d1115a0e3ed9abf41cb3ab80d19d622cbef7b93
Reviewed-on: https://chromium-review.googlesource.com/534594
Reviewed-by: Adam Klein <adamk@chromium.org>
Commit-Queue: Ben Smith <binji@chromium.org>
Cr-Commit-Position: refs/heads/master@{#45930}
[modify] https://crrev.com/284a4804f2f2cab172d2b4d96048b5d8986c72cd/src/bootstrapper.cc
[modify] https://crrev.com/284a4804f2f2cab172d2b4d96048b5d8986c72cd/src/contexts.h

OK, landed. I'll take a look at the perf charts in a day or two to see if this improved the regression.

OK, this change seems to have improved all of the memory graphs. The previous graph increased by about 23k, this change reduces it by 13k, for a net increase of ~10k. Unfortunately, there is a much larger memory regression ( https://crbug.com/729228 ), so the absolute ending memory usage is up by quite a bit more.

 Issue 724056  has been merged into this issue.

Issue 724969 has been merged into this issue.

This bug requires manual review: M60 has already been promoted to the beta branch, so this requires manual review
Please contact the milestone owner if you have questions.
Owners: amineer@(Android), cmasso@(iOS), josafat@(ChromeOS), bustamante@(Desktop)

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Please add appropriate OSs.  Thanks.

Rejected for M60 branch 3112 - it's not a huge regression, and the fix is not small.

Closing as fixed; binji, please reopen if I made a mistake?