All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=781550

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=86b89e6afa78178cc26b9e17f8c3aadfa2b53a56288894980a464c412b6a24ef


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/8963833485760684656

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

Hi hans@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 : Hans Wennborg
  Commit : c4a8c7b5ff68ad06bdaff91be5170e4da89c3a9f
  Date   : Fri Nov 03 05:28:21 2017
  Subject: Roll clang 315613:317263.

Bisect Details
  Configuration: winx64ati_perf_bisect
  Benchmark    : media_perftests
  Metric       : vector_math_fmac/optimized_aligned
  Change       : 8.77% | 585.112447256 -> 533.769283432

Revision             Result                  N
chromium@513642      585.112 +- 4.91672      6      good
chromium@513674      584.379 +- 2.70876      6      good
chromium@513690      586.068 +- 4.84439      6      good
chromium@513694      583.889 +- 2.84956      6      good
chromium@513696      585.481 +- 3.33382      6      good
chromium@513697      533.768 +- 3.24499      6      bad       <--
chromium@513698      531.194 +- 21.5025      6      bad
chromium@513705      533.769 +- 6.27193      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/8963833485760684656


For feedback, file a bug with component Speed>Bisection

This benchmark is a little silly becuase it exercises FMAC_SSE, but when building with Clang that function isn't actually used in Chrome, instead we use FMAC_C because Clang auto-vectorizes that to better code.

Anyway, the regression is very mysterious. I've looked at FMAC_SSE, and the machine code is identical between Clang versions. I've also looked at VectorMathPerfTest::RunPerfTest, and that is identical too.

I looked at vector_math_fmac/unoptimized, and the performance of that one hasn't regressed. Given that that's what we use in production, I'll close this as wontfix.