All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=714393

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgou6Q-gkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgorH3iwoM


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

linux-release

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8981591439910819936

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

Hi capn@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 : capn
  Commit : d85baf0b71c69bbd181aaefc8a803611e03c8eed
  Date   : Wed Apr 12 01:17:11 2017
  Subject: Enable SwiftShader fallback on desktop Linux.

Bisect Details
  Configuration: linux_perf_bisect
  Benchmark    : media.mse_cases
  Metric       : mediasource_opening/AV_buffer_post_load
  Change       : 2.96% | 14.9596428571 -> 17.4066666667

Revision             Result                  N
chromium@463837      14.9596 +- 70.8699      14      good
chromium@463863      9.51583 +- 8.02173      6       good
chromium@463870      9.84167 +- 7.25145      6       good
chromium@463871      9.29417 +- 8.87259      6       good
chromium@463872      17.4492 +- 7.30663      6       bad       <--
chromium@463873      17.7 +- 12.179          9       bad
chromium@463876      19.0833 +- 4.90984      6       bad
chromium@463889      16.9821 +- 10.3615      14      bad
chromium@463941      17.4067 +- 6.12116      9       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 media.mse_cases

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8981591439910819936

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5355610079494144


| 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 714394  has been merged into this issue.

I don't think my CL caused this. The reported regression is 3203%, while my change adds less than 100% which is only barely outside of the normal variability range.

For the record, my CL enabled using WebGL without a (whitelisted) GPU, by performing 3D rendering on the CPU. As far as I know Media Source Extensions should not depend on WebGL. But there is an off chance that something triggers the creation of the GPU process, even though it's not available for anything other than WebGL. In that case this would be related to  Issue 702268 .

The regression has disappeared since 4/22.