All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=722473

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg6tCs4wgM


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

chromium-rel-mac11

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8979505428759117376

https://chromeperf.appspot.com/group_report?rev=471046

Yes, this CL appears to have caused a large number of other perf regressions on Mac, and this bisect seems to have confirmed it for media_perftests. Will revert.

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

Hi pcc@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 : pcc
  Commit : ea5fad44dd0b0a74a40dbe2c05598f43861cd75a
  Date   : Thu May 11 19:42:44 2017
  Subject: Reland of build: Enable optimize_for_size unconditionally. (patchset #1 id:1 of https://codereview.chromium.org/2881503002/ )

Bisect Details
  Configuration: mac_10_11_perf_bisect
  Benchmark    : media_perftests
  Metric       : vector_math_fmac/unoptimized
  Change       : 72.40% | 697.672498722 -> 192.552001124

Revision             Result                   N
chromium@470967      697.672 +- 2.68955       6      good
chromium@471024      692.379 +- 17.8581       6      good
chromium@471038      703.407 +- 22.7848       6      good
chromium@471045      693.127 +- 59.0649       6      good
chromium@471046      192.906 +- 0.561239      6      bad       <--
chromium@471047      192.879 +- 1.16911       6      bad
chromium@471049      192.915 +- 0.631431      6      bad
chromium@471052      192.37 +- 0.683321       6      bad
chromium@471080      192.552 +- 1.27322       6      bad

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

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8979505428759117376

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5907368136671232


| 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!

https://codereview.chromium.org/2883113002/

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/22667180463b25a9059df17350b495b037e5a300

commit 22667180463b25a9059df17350b495b037e5a300
Author: pcc <pcc@chromium.org>
Date: Tue May 16 03:58:11 2017

Revert of build: Enable optimize_for_size unconditionally. (patchset #1 id:1 of https://codereview.chromium.org/2870393005/ )

Reason for revert:
Large number of perf regressions on mac.

BUG= 722473 

Original issue's description:
> Reland of build: Enable optimize_for_size unconditionally. (patchset #1 id:1 of https://codereview.chromium.org/2881503002/ )
>
> Reason for revert:
> Relanding now that the msan issue is fixed: https://codereview.chromium.org/2876693003
>
> Original issue's description:
> > Revert of build: Enable optimize_for_size unconditionally. (patchset #1 id:1 of https://codereview.chromium.org/2864383003/ )
> >
> > Reason for revert:
> > Causes failures on a MSAN bot.
> > https://uberchromegw.corp.google.com/i/chromium.memory/builders/Linux%20ChromiumOS%20MSan%20Tests/builds/440
> >
> >
> > Original issue's description:
> > > build: Enable optimize_for_size unconditionally.
> > >
> > > This change causes us to favor size over speed on Linux and Mac,
> > > which aligns the build config for those platforms with that of the
> > > other supported platforms, and should reduce the binary size impact
> > > of enabling ThinLTO. This change is expected to reduce binary size
> > > for Linux official builds by about 15%.
> > >
> > > There may be unacceptable perf regressions associated with this
> > > change, but the perf bots should at least let us know where those
> > > regressions are. I plan to monitor the Linux and Mac perf bots once
> > > it lands.
> > >
> > > BUG= 660216 
> > > R=thakis@chromium.org
> > >
> > > Review-Url: https://codereview.chromium.org/2864383003
> > > Cr-Commit-Position: refs/heads/master@{#470606}
> > > Committed: https://chromium.googlesource.com/chromium/src/+/c1269ce7fec8568a1789e07b2b5bb3b630a67f5f
> >
> > TBR=thakis@chromium.org,pcc@chromium.org
> > # Skipping CQ checks because original CL landed less than 1 days ago.
> > NOPRESUBMIT=true
> > NOTREECHECKS=true
> > NOTRY=true
> > BUG= 660216 
> >
> > Review-Url: https://codereview.chromium.org/2881503002
> > Cr-Commit-Position: refs/heads/master@{#470793}
> > Committed: https://chromium.googlesource.com/chromium/src/+/2a6f440fa84a584d916254f9a4c4e4d83c53435f
>
> TBR=thakis@chromium.org,yhirano@chromium.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG= 660216 
>
> Review-Url: https://codereview.chromium.org/2870393005
> Cr-Commit-Position: refs/heads/master@{#471046}
> Committed: https://chromium.googlesource.com/chromium/src/+/ea5fad44dd0b0a74a40dbe2c05598f43861cd75a

TBR=thakis@chromium.org,yhirano@chromium.org
# Not skipping CQ checks because original CL landed more than 1 days ago.
BUG= 660216 

Review-Url: https://codereview.chromium.org/2883113002
Cr-Commit-Position: refs/heads/master@{#472010}

[modify] https://crrev.com/22667180463b25a9059df17350b495b037e5a300/build/config/compiler/BUILD.gn

 Issue 722875  has been merged into this issue.

Perf recovered after my revert landed, e.g. https://chromeperf.appspot.com/report?sid=e8a4a1dde8d15c5cbe41f0385d97ec3d66e62147a30ce222608359fcd9954f1e&rev=471062

 Issue 724075  has been merged into this issue.

We don't care about regressions in unoptimized variants it's just there as a baseline for the SSE/NEON variants that we have. I see there were other regressions though, since these are mostly synthetic benchmarks (running code at absurd rates) did we see any power regressions?