All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=792452

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=e716ba9e9e0fc61e505bbb0b0c4c5f130bbca39db1216dca26bd43bf3f71c641


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

android-nexus5
android-nexus6
android-webview-nexus5X
android-webview-nexus6

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8960949412425707456

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

Hi chrishtr@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 : Chris Harrelson
  Commit : a3e4477c29d087d3708dbcb4dc5a90985c2b45e3
  Date   : Tue Dec 05 01:51:58 2017
  Subject: [PE] Turn on half-width tiles for all platforms and devices.

Bisect Details
  Configuration: android_webview_nexus6_aosp_perf_bisect
  Benchmark    : thread_times.simple_mobile_sites
  Metric       : thread_GPU_cpu_time_per_frame/http___www.nyc.gov
  Change       : 27.19% | 1.06848847932 -> 1.35896728206

Revision             Result                    N
chromium@521570      1.06849 +- 0.0557503      6      good
chromium@521581      1.07507 +- 0.051197       6      good
chromium@521584      1.0813 +- 0.0695394       6      good
chromium@521585      1.05921 +- 0.0479745      6      good
chromium@521586      1.35518 +- 0.0744642      6      bad       <--
chromium@521591      1.36833 +- 0.0691084      6      bad
chromium@521612      1.35897 +- 0.074822       6      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-webview --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=http...www.nyc.gov thread_times.simple_mobile_sites

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

Debug information about this bisect:
  https://chromeperf.appspot.com/buildbucket_job_status/8960949412425707456


For feedback, file a bug with component Speed>Bisection

 Issue 792448  has been merged into this issue.

 Issue 792454  has been merged into this issue.

 Issue 792455  has been merged into this issue.

 Issue 792450  has been merged into this issue.

 Issue 792451  has been merged into this issue.

 Issue 792460  has been merged into this issue.

 Issue 792456  has been merged into this issue.

 Issue 792909  has been merged into this issue.

 Issue 792453  has been merged into this issue.

 Issue 792915  has been merged into this issue.

 Issue 792459  has been merged into this issue.

 Issue 792940  has been merged into this issue.

 Issue 798608  has been merged into this issue.

 Issue 798824  has been merged into this issue.

 Issue 792458  has been merged into this issue.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/7db283d1d769d05e88b565d5b583980f3d2202aa

commit 7db283d1d769d05e88b565d5b583980f3d2202aa
Author: Eric Karl <ericrk@chromium.org>
Date: Wed Jan 10 21:35:18 2018

GPU raster tile size calculation should account for DSF rounding

On certain device screens / device scale factors (DSF), we can end up with non-optimal
tilings as GPU raster does not take DSF rounding into account when calculating tile width.

Consider a theoretical device with a screen resolution of 1502x2560 and a DSF of 3.5.

In this case, mobile content with 100% width would be sized as:
ceil(1502 / 3.5) = ceil(429.1...) = 430 DIPs

Because of the ceil operation, we've grown the content bounds slightly, and if we convert
the content bounds back to device pixels (DPs) we get:

ceil(430 * 3.5) = 1505 DPs

Despite content being sized at 1505, GPU tile width is based on original viewport width
(1502). With rounding, this can sometimes work out, but in this case we end up with a
final GPU tile size of 1504.

Because the tile is narrower than the content, we end up using an entire extra tile to
accommodate 1 column of pixels.

Instead, we should adjust the GPU tile size for the DSF rounding, avoiding this case.

R=vmpstr

Bug:  800604 ,  792452 
Cq-Include-Trybots: master.tryserver.blink:linux_trusty_blink_rel;master.tryserver.chromium.android:android_optional_gpu_tests_rel
Change-Id: Ia39025374d3d30ebbf5e59508353648524f627f5
Reviewed-on: https://chromium-review.googlesource.com/858483
Reviewed-by: vmpstr <vmpstr@chromium.org>
Commit-Queue: Eric Karl <ericrk@chromium.org>
Cr-Commit-Position: refs/heads/master@{#528431}
[modify] https://crrev.com/7db283d1d769d05e88b565d5b583980f3d2202aa/cc/layers/picture_layer_impl.cc
[modify] https://crrev.com/7db283d1d769d05e88b565d5b583980f3d2202aa/cc/layers/picture_layer_impl_unittest.cc

With the above change, all cases where we don't truly have wider-than-viewport content have returned to previous values. For those cases with wider-than-viewport content,
this is an intentional change which should reduce memory and may reduce prepaint, at
the cost of a few more tiles.

 Issue 798613  has been merged into this issue.

 Issue 804316  has been merged into this issue.