All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=737350

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg7tHgogoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgruj3uwoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgrt_XrQkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg7tjFtgkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg7qj2ogoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgrrHIogoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg7tL9ugkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgroj46QgM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgrsr_oQoM


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

android-nexus5
android-nexus6
android-webview-nexus6

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8975584241288338720

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

Hi mstarzinger@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 : Michael Starzinger
  Commit : 0d833cb94f882e28054ea1fcff7de1d6dafe302a
  Date   : Mon Jun 26 15:04:43 2017
  Subject: [deoptimizer] Remove support for code-stub "deopt".

Bisect Details
  Configuration: android_webview_nexus6_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       : 1.58% | 2015572.0 -> 2047372.0

Revision                           Result                  N
chromium@482301                    2015572 +- 0.0          6      good
chromium@482334                    2015572 +- 0.0          6      good
chromium@482342                    2015588 +- 87.6356      6      good
chromium@482344                    2015572 +- 0.0          6      good
chromium@482345                    2015572 +- 0.0          6      good
chromium@482345,v8@d4b4b7e35f      2015572 +- 0.0          6      good
chromium@482345,v8@0d833cb94f      2047372 +- 0.0          6      bad       <--
chromium@482345,v8@acf4929379      2047372 +- 0.0          6      bad
chromium@482345,v8@71eb30bcdd      2047372 +- 0.0          6      bad
chromium@482346                    2047372 +- 0.0          6      bad
chromium@482350                    2047372 +- 0.0          6      bad
chromium@482366                    2047372 +- 0.0          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/8975584241288338720

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=6624853068611584


| 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 737488  has been merged into this issue.

The CL in question just removed dead code. If at all then memory footprint should have gone down, not up. There must be some secondary effect that is causing this regression.

ulan, hpayer: any ideas how to debug what's happening here?

Michael Lippautz helped me investigate this yesterday. Our conclusion was that the removal of the two stubs (i.e. the two StubFailureTrampolineStub instances) changed GC timing so that there is one less GC happening. This is supported by the size of large-object space being bigger on some benchmarks after this CL. Unfortunately none of the affected benchmarks has a "GC count" metric, which would fully validate our theory.

+perezju, hjd, primiano: see #8, does the gc timing theory make sense? any way to verify?

WontFix-ing since there don't seem to be objections to the theory in #8