New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 678572 link

Starred by 1 user

Issue metadata

Status: WontFix
Owner:
Last visit > 30 days ago
Closed: Jan 2017
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

64.3% regression in memory.top_10_mobile at 441202:441231

Project Member Reported by primiano@chromium.org, Jan 5 2017

Issue description

I  suspect https://codereview.chromium.org/2605273003, let's see if bisect says something
 
All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=678572

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg__PkpgoM


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

android-nexus5X
Cc: ericrk@chromium.org
Owner: ericrk@chromium.org

=== PERF REGRESSION ===


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

Hi ericrk@chromium.org, the bisect results pointed to your CL, please take a look at the
results.


===== BISECT JOB RESULTS =====
Status: completed


===== SUSPECTED CL(s) =====
Subject : Don't use DF text during animations
Author  : ericrk
Commit description:
  
Previously, we enabled distance-field text during animations in
order to improve GPU raster performance. Now that GPU raster
handles animations the same way as SW - only rastering at specific
scales and interpolating, this optimization isn't necessary.

In certain cases, the optimization causes visual differences
between non-animating and animating content, so removing this
improves our correctness as well.

R=ajuma@chromium.org
BUG= 668723 
CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_trusty_blink_rel

Review-Url: https://codereview.chromium.org/2605273003
Cr-Commit-Position: refs/heads/master@{#441204}
Commit  : b04a4a62bc244a238e63c6862be0ed6dfe84a837
Date    : Tue Jan 03 20:51:04 2017


===== TESTED REVISIONS =====
Revision         Mean    Std Dev  N  Good?
chromium@441201  151985  4660.37  6  good
chromium@441203  151313  3684.35  6  good
chromium@441204  249129  2263.58  6  bad    <--
chromium@441205  249307  2669.9   6  bad
chromium@441209  249532  2421.72  6  bad
chromium@441216  248679  1409.67  6  bad
chromium@441231  249259  2648.09  6  bad

Bisect job ran on: android_nexus5X_perf_bisect
Bug ID: 678572

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=https.www.google.co.uk.hl.en.q.science memory.top_10_mobile
Test Metric: memory:chrome:all_processes:reported_by_chrome:skia:effective_size_avg/foreground/https_www_google_co_uk_hl_en_q_science
Relative Change: 64.00%

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus5X_perf_bisect/builds/1035
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8991308352342113488


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5821517856768000

| 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 Tests>AutoBisect.  Thank you!

Comment 4 by ericrk@chromium.org, Jan 11 2017

Owner: bsalomon@chromium.org
Status: Assigned (was: Untriaged)
I think this is WontFix:
We switched from using DF to normal text for animating layers. This causes slightly more memory usage in the test in question (150kb to 250kb). This change comes from the GlyphCache. My guess is that we are creating more glyphs than before, as we could re-use the DF glyphs at more scales.

Since we made this change for correctness, and the delta is small, I think we should just close this out. bsalomon@, do you agree?
Status: WontFix (was: Assigned)
sgtm

Sign in to add a comment