All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=690209

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgqLu7qgkM


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

chromium-rel-mac11

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8988183247901220384

=== Auto-CCing suspected CL author sigbjornf@opera.com ===

Hi sigbjornf@opera.com, 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 : sigbjornf
  Commit : 05cea29b1f7245b2cbd42f7da77705b9fa820707
  Date   : Thu Feb 02 01:56:23 2017
  Subject: Have SubframeLoadingDisabler singleton live off-heap.

Bisect Details
  Configuration: mac_10_11_perf_bisect
  Benchmark    : blink_perf.layout
  Metric       : nested-grid/nested-grid
  Change       : 5.45% | 299.055776479 -> 282.748965143

Revision             Result                  N
chromium@447568      299.056 +- 4.86698      6      good
chromium@447632      291.989 +- 2.61943      6      good
chromium@447664      288.764 +- 4.00186      6      good
chromium@447680      302.817 +- 1.8371       6      good
chromium@447681      302.141 +- 4.9252       6      good
chromium@447682      280.7 +- 4.43606        6      bad       <--
chromium@447684      281.02 +- 2.89914       6      bad
chromium@447688      277.801 +- 5.78528      6      bad
chromium@447696      282.749 +- 2.98498      6      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests blink_perf.layout

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8988183247901220384

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=4926079942262784


| 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 Tests>AutoBisect.  Thank you!

That CL addresses a stability issue by moving the allocation of a singleton map from one allocator to another, the singleton tracking a set of root documents.

It makes little sense for that to have a bearing on a "display:grid" performance test, other than the changed allocation having a non-local effect on memory subsystem performance which is somehow coming through. None of the nearby commits could otherwise explain the drop, afaict.

+eae

Emil: looks like this regression is really only about 5%, and only on one of our mac configs (a mac mini running 10.11). Based on that and #4, do you think we should pursue a fix?

Probably not worth it, assuming it only affects a single config. The nested test perf test is't terribly important.