All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=721904

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgqoC5-AgM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgyuv8vQkM


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

chromium-rel-mac11

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8979767580732945584

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

Hi fdoray@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 : fdoray
  Commit : e324edd539822b440ee78bd375f8dc0e284f39b6
  Date   : Mon May 08 16:21:46 2017
  Subject: Remove usage of ScopedTaskScheduler.

Bisect Details
  Configuration: mac_10_11_perf_bisect
  Benchmark    : media_perftests
  Metric       : demuxer_bench/bear.flac
  Change       : 34.55% | 1697.43298179 -> 1111.04459549

Revision             Result                  N
chromium@469997      1697.43 +- 38.2808      6      good
chromium@470002      1677.47 +- 26.7939      6      good
chromium@470003      1106.94 +- 12.4593      6      bad       <--
chromium@470004      1105.98 +- 18.166       6      bad
chromium@470005      1125.11 +- 53.15        6      bad
chromium@470007      1106.85 +- 9.88799      6      bad
chromium@470016      1115.98 +- 28.9399      6      bad
chromium@470035      1119.22 +- 39.4321      6      bad
chromium@470072      1113.66 +- 19.5128      6      bad
chromium@470146      1111.04 +- 10.8927      6      bad

To Run This Test
  ./src/out/Release/media_perftests --single-process-tests

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8979767580732945584

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=4968216528420864


| 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 721895  has been merged into this issue.

The suspected CL replaces a ScopedTaskScheduler (deprecated) with a ScopedTaskEnvironment. A ScopedTaskScheduler runs tasks posted via base/task_scheduler/post_task.h on the main thread while ScopedTaskEnvironment runs them on separate threads. The behavior of ScopedTaskEnvironment is closer from what users get. Is it possible that the main thread is able to run tasks faster than other threads? (erikchen@: On Mac, thread priority is higher on the main thread by default?)

The graphs show that chromium-rel-mac11/media_perftests/demuxer_bench/bear.flac improved in the range r454364 - r454510, which is when FFmpegDemuxer started to post tasks to TaskScheduler  r454452 Therefore, this isn't a real regression but rather a return to the numbers that we had before starting to post to TaskScheduler.

Seems fine since this is a test only change. It'd be nice if the perf test used the same environment as production, but it's not required here.