All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=710017

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgvMjB-gkM


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

chromium-rel-win7-gpu-nvidia

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8982684591535580336

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

Hi ckrasic@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 : ckrasic
  Commit : dee375739b384173dea843b8f4e1632a4318d4d3
  Date   : Thu Apr 06 22:42:27 2017
  Subject: QUIC - fix crash bug in client handling of Server Push.

Bisect Details
  Configuration: winx64nvidia_perf_bisect
  Benchmark    : media_perftests
  Metric       : audio_converter/convert_pass_through
  Change       : 2.03% | 2083934.07647 -> 2041694.86687

Revision             Result                  N
chromium@462653      2083934 +- 13533.6      6      good
chromium@462656      2084529 +- 19471.3      6      good
chromium@462658      2069779 +- 16266.9      6      good
chromium@462659      2041855 +- 54067.8      6      bad       <--
chromium@462665      2044788 +- 48595.1      6      bad
chromium@462677      2041634 +- 16720.7      6      bad
chromium@462701      2019107 +- 17838.6      6      bad
chromium@462748      2029604 +- 40463.5      6      bad
chromium@462843      2041695 +- 26902.7      6      bad

To Run This Test
  .\src\out\Release_x64\media_perftests.exe --single-process-tests

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8982684591535580336

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5824185563086848


| 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!

The bisect must be wrong.  No way this change should effect a media related metric.

Yeah, you're right. This shouldn't be due to your change. I suspect something is flaky in bit-alignment or something else.

Dale, is this metric important enough that we should investigate with ETW traces?

This metric is kind of important; it covers how efficient our audio conversion (mixing, resampling, etc) pipeline is. It's used by many components, so if this is a legit (non-recovering) regression we should try to track it down.

Looking into this now.

I can't repro a regression locally. Here's my data:

good 462653 6e2e787479f591d24386ba91d8bbf3d1cfbe1fd3
1.  1789997.4940035085 runs/s
2.  1832189.7415696369 runs/s
3.  1835519.1306981395 runs/s
4.  1793947.2220727266 runs/s
5.  1809087.0442231328 runs/s
6.  1803377.7264817 runs/s


bad 462843 539e731c557cbce418d3ac1ea315f6cc9bd0e21f
1.  1775836.1968692008 runs/s
2.  1745459.6231552672 runs/s
3.  1825600.3943296853 runs/s
4.  1796848.3280326307 runs/s
5.  1732336.6623069528 runs/s
6.  1762642.553716532 runs/s
7.  1822174.0358421633 runs/s
8.  1814322.259919807 runs/s
9.  1839672.5382881847 runs/s

Probably this is just caused by a change in the bot. I filed crbug/714812 to see about getting reference builds working for these tests.

 Issue 713123  has been merged into this issue.