All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=783650

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=01d657ab2d64b69ba6510046d110fe2f2f1df58f1ef8b44877d604b57c0296c8


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

chromium-rel-mac12
chromium-rel-win10

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8963323204934066704

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

Hi zmo@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 : Zhenyao Mo
  Commit : 1202f66559ca7e37f643a691cac72144ab60493f
  Date   : Thu Nov 09 00:45:05 2017
  Subject: [reland] Move GPU compositing blacklisting to GPU process.

Bisect Details
  Configuration: winx64_10_perf_bisect
  Benchmark    : memory.desktop
  Metric       : memory:chrome:all_processes:reported_by_chrome:malloc:effective_size_avg/TrivialScrollingPageSharedPageState
  Change       : 3.91% | 12022409.0 -> 12492164.5

Revision             Result                   N
chromium@515008      12022409 +- 18174.5      6      good
chromium@515032      12030823 +- 35499.4      6      good
chromium@515035      12037005 +- 15121.4      6      good
chromium@515036      12457466 +- 56521.6      6      bad       <--
chromium@515037      12488879 +- 49929.1      6      bad
chromium@515038      12459386 +- 81668.2      6      bad
chromium@515044      12461875 +- 74248.5      6      bad
chromium@515055      12462957 +- 50871.3      6      bad
chromium@515101      12492165 +- 69896.4      6      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=TrivialScrollingPageSharedPageState memory.desktop

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8963323204934066704


For feedback, file a bug with component Speed>Bisection

 Issue 783722  has been merged into this issue.

 Issue 783723  has been merged into this issue.

 Issue 783721  has been merged into this issue.

 Issue 783697  has been merged into this issue.

 Issue 783870  has been merged into this issue.

 Issue 783869  has been merged into this issue.

Looking at both win10_perf and win8_perf bots, they are running with GPU compositing blacklisted.  This is not the path we ship to most users and not the path we optimize.

With my CL, we delay the blacklist decision to until GPU process launch and return a GPU channel, so we end up doing some extra wasted work if GPU compositing is blacklisted. This is expected.

We'll have to live with the new baselines for these tests.

However, if we run perf tests on bots with real GPU and GPU compositing enabled, such regression won't happen.