All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=766204

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=57ffec47b985f713b14b37802363b0f5c119245a0ab8cb65a63e709c05daf8d3


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

chromium-rel-win10
chromium-rel-win7-gpu-ati
chromium-rel-win7-gpu-nvidia
chromium-rel-win8-dual

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8968097525382857824

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

Hi brucedawson@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 : Bruce Dawson
  Commit : 8707de2d756841d799483ceff0427e316961f468
  Date   : Fri Sep 15 23:00:10 2017
  Subject: Make VS 2017 the default compiler on Windows

Bisect Details
  Configuration: winx64_10_perf_bisect
  Benchmark    : media_perftests
  Metric       : sinc_resampler_convolve/unoptimized_aligned
  Change       : 12.76% | 36374.0580357 -> 31730.926812

Revision             Result                  N
chromium@502407      36374.1 +- 26.6764      6      good
chromium@502417      36382.7 +- 18.3386      6      good
chromium@502420      35730.4 +- 784.492      6      good
chromium@502421      32193.9 +- 101.168      6      bad       <--
chromium@502422      31109.6 +- 1832.72      6      bad
chromium@502426      31872.2 +- 60.3502      6      bad
chromium@502445      31884.8 +- 39.2661      6      bad
chromium@502483      31730.9 +- 1246.65      6      bad

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

More information on addressing performance regressions:
  http://g.co/ChromePerformanceRegressions

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8968097525382857824


For feedback, file a bug with component Speed>Bisection

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8968084339254979408

Yep, known issue. Code-gen changes. They *shouldn't* have that big an impact on this micro-benchmark, but they do. VS 2017 has now been reverted but we will hit this again when we finally switch. At that point I believe we will "Won't Fix" that bug, which is what I am going to do with this bug now, because it's not enough to block VS 2017 and there is no known workaround.

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

Hi fs@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 : Fredrik Söderquist
  Commit : f5eb27c2b897f206b275fd862e874b64159cc15e
  Date   : Fri Sep 15 22:16:06 2017
  Subject: Enable Skia's SkImageGenerator implementation

Bisect Details
  Configuration: win_8_perf_bisect
  Benchmark    : media_perftests
  Metric       : clockless_playback/sfx.flac
  Change       : 47.68% | 3991.97738968 -> 5895.28289576

Revision             Result                  N
chromium@502391      3991.98 +- 743.786      6      good
chromium@502397      4389.33 +- 754.903      6      good
chromium@502400      3469.92 +- 411.941      6      good
chromium@502401      7471.68 +- 1134.85      6      bad       <--
chromium@502402      7167.62 +- 909.762      6      bad
chromium@502403      7074.71 +- 1241.62      6      bad
chromium@502414      6700.65 +- 1202.73      6      bad
chromium@502436      6335.87 +- 1342.93      6      bad
chromium@502481      5895.28 +- 1300.98      6      bad

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

More information on addressing performance regressions:
  http://g.co/ChromePerformanceRegressions

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8968084339254979408


For feedback, file a bug with component Speed>Bisection

Be careful to lock down to a particular compiler while investigating since we changed to VS 2017 and then back to VS 2015 around this time range and it affects some tests (sinc_resampler_convolve/unoptimized_aligned). Forcing to VS 2015 with set GYP_MSVS_VERSION=2015 & gclient runhooks & gn gen out\yourdir is probably best to avoid complications.

It looks like the indicated test (and all the similar ones in the report) have recovered, which would seem to be a strong enough indication that the CL in c#6 was not the culprit. (Also as a first impression it seemed odd that a change to image decoding would have an effect on something relating to audio decoding...)