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

Issue 618676 link

Starred by 2 users

Issue metadata

Status: Duplicate
Merged: issue 618679
Owner:
Last visit > 30 days ago
Closed: Jul 2016
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

1.2%-62.6% regression in smoothness.gpu_rasterization_and_decoding.image_decoding_cases at 397457:398066

Project Member Reported by pmeenan@chromium.org, Jun 9 2016

Issue description

See the link to graphs below.
 
All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=618676

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgvLattAkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgvNGwuAoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3OO8qQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgvNjvqgoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgvOy8sQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3JG4rAkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3PukoAkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3OOmuAoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3M7HuQkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3Kr3sQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgvNaTvwoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3OPosAoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgvMqpvgkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3MWNvgkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgvJaJvQsM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3KOFrgsM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3I7wvAkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3N-qtgoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3J-JowkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgvLi4uwoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3ImooAoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3KXDrAoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg3IesvQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgvOCEuQoM


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

chromium-rel-win10
chromium-rel-win7-dual
chromium-rel-win7-gpu-ati
chromium-rel-win7-gpu-intel
chromium-rel-win7-gpu-nvidia
chromium-rel-win7-single
chromium-rel-win7-x64-dual
chromium-rel-win8-dual
Cc: dpranke@chromium.org
Owner: dpranke@chromium.org

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

Hi dpranke@chromium.org, the bisect results pointed to your CL below as possibly
causing a regression. Please have a look at this info and see whether
your CL be related.


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


===== SUSPECTED CL(s) =====
Subject : Flip Linux bots on chromium.lkgr to GN.
Author  : dpranke
Commit description:
  
R=brettw@chromium.org, inferno@chromium.org
BUG= 605732 

Review-Url: https://codereview.chromium.org/2006173002
Cr-Commit-Position: refs/heads/master@{#398038}
Commit  : 6864dbb0407a47f697901f70105d2c2c829c08bc
Date    : Mon Jun 06 16:05:51 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@398032  45.2673  6.38707  18  good
chromium@398035  48.9372  7.49998  18  good
chromium@398037  43.2888  2.96168  5   good
chromium@398038  56.7873  2.42553  12  bad    <--

Bisect job ran on: win_perf_bisect
Bug ID: 618676

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests smoothness.top_25_smooth
Test Metric: mean_input_event_latency/https___mail.google.com_mail_
Relative Change: 28.65%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/win_perf_bisect/builds/6581
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9010327844000194848


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5325100564348928

| 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!
Owner: pmeenan@chromium.org
That CL wouldn't have affected those bots. Can we re-run the bisect again?
Re-kicked with the Docs latency test which is a much cleaner regression to bisect.
Kicked bisect again with a narrower range from http://test-results.appspot.com/revision_range?start=398031&end=398034 to see if it was from the V8 roll and to see if it can pick the offender within that.

New bisect: https://chromeperf.appspot.com/buildbucket_job_status/9008674556393241728
Project Member

Comment 6 by sheriffbot@chromium.org, Jul 6 2016

Labels: -M-53 M-54 MovedFrom-53
Moving this nonessential bug to the next milestone.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
Mergedinto: 618679
Status: Duplicate (was: Assigned)

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


===== SUSPECTED CL(s) =====
Subject : Making cookies eviction quotas match spec
Author  : maksim.sisov
Commit description:
  
According to the section 3 and section 5 in
https://tools.ietf.org/html/draft-west-cookie-priority-
00#section-3, the old cookie monster implementation doesn't
maintain priority quotas correctly.

This CL makes sure there will always be at least 30 low
priority cookies, 50 mid priority and 70 high priority
cookies if there had been enough of them before  the eviction.
Please note, secure cookies are more important than non-secure
per https://tools.ietf.org/html/draft-west-leave-secure-cookies-alone.

A small example: if there are 70 cookies:
37 non-secure low priority and 33 secure low priority
cookies, 37 non-secure cookies are deleted during the first
round and other 3 secure low priority cookies are deleted
during the following round preserving 30 low priority
cookies according to the quota of those specific cookies.
For a bigger example that fully complies with the
implementation, check the unittests.

Before the fix, the unittests were just adjusted to the
behavior of cookies eviction.

For example, if we take the following unittest:
Round 1 => 10L; round 2 => 11M, 10L; round 3 => none.
TestPriorityCookieCase(cm.get(), "11HN 10MN 20LN 110MN 20LN
10HN", 20U,109U, 21U, 150U, 0U);
The problem here was that there were only 40 low priority
cookies, but the quota was not preserved for those cookies.
First, 10 low priority cookies were deleted and then more 10
low priority cookies were deleted leaving only 20 of them,
which was less than the quota (30 low priority cookies).
It happened because the eviction algorithm didn't know how
many cookies of a specific priority were deleted and
it had always started to delete all the cookies from the
beginning of the container removing even those cookies
that shouldn't have been deleted.

After we land this CL, we can have cookies in any order and
high priority cookies will be eventually deleted in order
to avoid excess of high priority cookies by some
applications within the same domain. Thus, after the
eviction algorithm runs, we should have at least 30 low, 50
mid and 70 high priority cookies if we had sufficient
amount of them in the beginning.

BUG= 609550 

Review-Url: https://codereview.chromium.org/1976073002
Cr-Commit-Position: refs/heads/master@{#398049}
Commit  : fe1fea6df040fff38a3a4567e20a3d5847e83170
Date    : Mon Jun 06 16:53:15 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev   N  Good?
chromium@398014  22.3402  0.428997  5  good
chromium@398032  22.4269  0.25292   5  good
chromium@398041  22.1335  0.438768  5  good
chromium@398046  22.4392  0.352764  5  good
chromium@398048  21.8324  0.343497  5  good
chromium@398049  23.2827  0.366862  5  bad    <--
chromium@398050  23.6673  0.147831  5  bad

Bisect job ran on: win_perf_bisect
Bug ID: 618676

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests smoothness.key_desktop_move_cases
Test Metric: mean_frame_time/mean_frame_time
Relative Change: 5.94%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/win_perf_bisect/builds/6656
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9007881165646610768


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5250327645782016

| 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!

Sign in to add a comment