Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8970870438470653760

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8970869926919194304

I recon this will bisect to https://chromium.googlesource.com/chromium/src/+/dc5e1afbd1f6b01e44b58ee16f4c556d2ceaa04c

... and maybe partly recovered at https://chromium.googlesource.com/chromium/src/+/4b44a058abace01b33f1b75798274dd07db13ddb

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

Hi vmpstr@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 : vmpstr
  Commit : dc5e1afbd1f6b01e44b58ee16f4c556d2ceaa04c
  Date   : Mon May 15 19:59:52 2017
  Subject: cc: Approximate ideal contents scale under perspective transform.

Bisect Details
  Configuration: android_nexus5_perf_bisect
  Benchmark    : thread_times.tough_compositor_cases
  Metric       : thread_GPU_cpu_time_per_frame/http___jsbin.com_giqafofe_1_quiet?JS_POSTER_CIRCLE
  Change       : 19.53% | 8.45047681911 -> 10.1008768423

Revision             Result                   N
chromium@471583      8.45048 +- 0.197169      6      good
chromium@471798      8.50313 +- 0.204401      6      good
chromium@471852      8.42865 +- 0.198817      6      good
chromium@471866      8.43259 +- 0.145705      6      good
chromium@471873      8.32129 +- 0.166501      6      good
chromium@471875      8.2589 +- 0.152146       6      good
chromium@471876      9.93818 +- 0.202855      6      bad       <--
chromium@471879      10.0767 +- 0.309449      6      bad
chromium@471905      10.1349 +- 0.194039      6      bad
chromium@472012      10.1009 +- 0.146143      6      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=http...jsbin.com.giqafofe.1.quiet.JS.POSTER.CIRCLE thread_times.tough_compositor_cases

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8970870438470653760


For feedback, file a bug with component Speed>Bisection

=== BISECT JOB RESULTS ===
Perf regression found with culprit

Suspected Commit
  Author : vmpstr
  Commit : dc5e1afbd1f6b01e44b58ee16f4c556d2ceaa04c
  Date   : Mon May 15 19:59:52 2017
  Subject: cc: Approximate ideal contents scale under perspective transform.

Bisect Details
  Configuration: android_nexus5_perf_bisect
  Benchmark    : thread_times.tough_compositor_cases
  Metric       : thread_raster_cpu_time_per_frame/http___jsbin.com_giqafofe_1_quiet?JS_POSTER_CIRCLE
  Change       : 390.53% | 0.0835665077527 -> 0.4099188982

Revision             Result                       N
chromium@471743      0.0835665 +- 0.00541099      6      good
chromium@471811      0.087946 +- 0.0104529        6      good
chromium@471845      0.0839644 +- 0.00857538      6      good
chromium@471862      0.0830926 +- 0.0172279       6      good
chromium@471870      0.0864394 +- 0.0109042       6      good
chromium@471874      0.0861682 +- 0.00529871      6      good
chromium@471875      0.0847007 +- 0.0153124       6      good
chromium@471876      0.411113 +- 0.0215564        6      bad       <--
chromium@471878      0.409189 +- 0.0175897        6      bad
chromium@472012      0.409919 +- 0.0302628        6      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=http...jsbin.com.giqafofe.1.quiet.JS.POSTER.CIRCLE thread_times.tough_compositor_cases

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8970869926919194304


For feedback, file a bug with component Speed>Bisection

The patch in question improves the quality of the page under perpsective transforms (see  issue 320369 ). #4 analysis is correct, this would regress performance since it is rasterizing higher quality content, and the second patch limits the impact to not have overly large scales. 

I'm leaning towards WontFix here. Please reopen if you think this isn't acceptable.

Looks reasonable to me.  Thanks!