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

Issue 655140 link

Starred by 2 users

Issue metadata

Status: WontFix
Owner:
Closed: Aug 2017
Cc:
EstimatedDays: ----
NextAction: 2016-10-20
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

10.9% regression in dromaeo.domcorequery at 424013:424071

Project Member Reported by tdres...@chromium.org, Oct 12 2016

Issue description

See the link to graphs below.
 
All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=655140

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgjcOS6woM


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

android-nexus7v2
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Oct 12 2016

Cc: ma...@chromium.org
Owner: ma...@chromium.org

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

Hi mathp@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 : [Autofill] Log count histogram for the number of server cards
Author  : mathp
Commit description:
  
BUG= 653636 
TEST=AutofillMetrics

Review-Url: https://codereview.chromium.org/2395243002
Cr-Commit-Position: refs/heads/master@{#424053}
Commit  : 2af1bc9bc29520e5b3893b8c31d008d1b384b296
Date    : Sat Oct 08 03:04:41 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N  Good?
chromium@424012  2880.43  113.447  5  good
chromium@424042  2857.93  50.1121  5  good
chromium@424050  2890.96  57.6191  5  good
chromium@424052  2853.33  86.2051  5  good
chromium@424053  2668.79  54.033   5  bad    <--
chromium@424054  2650.38  95.0233  5  bad
chromium@424057  2424.93  38.259   5  bad
chromium@424071  2477.14  68.0115  5  bad

Bisect job ran on: android_nexus7_perf_bisect
Bug ID: 655140

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests dromaeo.domcorequery
Test Metric: dom/dom
Relative Change: 14.00%
Score: 99.0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus7_perf_bisect/builds/3389
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8999003206035240624


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5842791261798400

| 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 ma...@chromium.org, Oct 12 2016

Hi, I don't think logging two histograms will incur a 10% regression. Is it possible another change is to blame? Thanks!

Comment 5 by ma...@chromium.org, Oct 13 2016

Owner: tdres...@chromium.org
Project Member

Comment 7 by 42576172...@developer.gserviceaccount.com, Oct 13 2016

Cc: ssid@chromium.org
Owner: ssid@chromium.org

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

Hi ssid@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 : Add LevelDB database names to tracing path for easier debugging
Author  : ssid
Commit description:
  
The data from field trials show large databases, but not sure what are
their names. So, adding names to the path.

BUG= 645126 

Review-Url: https://codereview.chromium.org/2399903004
Cr-Commit-Position: refs/heads/master@{#424055}
Commit  : 279f8d702e8b2fffda1b7f50b311c5202331e121
Date    : Sat Oct 08 03:21:14 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N  Good?
chromium@424012  2864.61  85.3338  5  good
chromium@424042  2842.83  60.3292  5  good
chromium@424050  2888.03  54.8605  5  good
chromium@424054  2722.15  43.6809  5  good
chromium@424055  2466.06  57.904   5  bad    <--
chromium@424056  2452.38  31.5053  5  bad
chromium@424057  2445.74  41.0923  5  bad
chromium@424071  2465.7   35.4628  5  bad

Bisect job ran on: android_nexus7_perf_bisect
Bug ID: 655140

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests dromaeo.domcorequery
Test Metric: dom/dom
Relative Change: 13.93%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus7_perf_bisect/builds/3393
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8998905668595918048


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5846651900526592

| 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 8 by ssid@chromium.org, Oct 13 2016

Cc: iclell...@chromium.org
Owner: dglazkov@chromium.org
This is a V8 specific benchmark. My change was in browser and in tracing. So, it should not affect.

I don't have much context about V8. I think dglazkov@ change could have cause this regression.

V8 related CLs:

commit	28e4cd6cdb271055f445717508ebea7595fa4676	
author	dglazkov <dglazkov@chromium.org>	Sat Oct 08 04:51:00 2016
committer	Commit bot <commit-bot@chromium.org>	Sat Oct 08 04:52:49 2016
Move the setting of bindings code generator up the stack.

This opens the path for cleanly switching different types of
code generators. Also eliminates the need for IdlCompilerV8 and
IdlCompilerDictionaryImpl.

R=bashi,yukishiino
BUG= 654129 

Review-Url: https://codereview.chromium.org/2401043003
Cr-Commit-Position: refs/heads/master@{#424070}


commit	a420fe39c98570d80347675134d3738af20c63b7	
author	iclelland <iclelland@chromium.org>	Sat Oct 08 01:10:00 2016
committer	Commit bot <commit-bot@chromium.org>	Sat Oct 08 01:12:31 2016
[Origin Trials] Clarify the distinction between trials and features

There is some confusion in the origin trial method names and comments
between the name of a feature, which is declared in
RuntimeEnabledFeatures, and a trial, which can enable multiple features.

This CL attempts to make a distinction between the two, most visibly by
renaming OriginTrialContext::isFeatureEnabled to
OriginTrialContext::isTrialEnabled.

R=chasej@chromium.org

Review-Url: https://codereview.chromium.org/2398793003
Cr-Commit-Position: refs/heads/master@{#424033}

Cc: -iclell...@chromium.org dglazkov@chromium.org
Owner: iclell...@chromium.org
ssid@ that's unlikely. My change didn't make any changes to the generated code. Ian, would you check and see if your change could be the culprit?
I don't think that's possible; my change (https://crrev.com/a420fe39c98570d80347675134d3738af20c63b7) only renamed one method, one instance var, and updated comments. Shouldn't have changed the binary at all.
Cc: -ssid@chromium.org
Owner: ssid@chromium.org
ssid@, any other candidates? What's the range? The bisect links are confusing.

Comment 13 by ssid@chromium.org, Oct 13 2016

Cc: dtu@chromium.org nednguyen@chromium.org
I do not think the bisects are going to work, the metric is very noisy. Having said that I am not even sure if this a real regression since the metric came back up to a good extent on the next runs.

The range of CLs that could be responsible:

https://chromium.googlesource.com/chromium/src/+log/29d66d91e305746076cadfc02065a908beeff129%5E..98e99ecd2fb3fbffbe6a622b0f38a3e930a37964?pretty=fuller

+dtu, Ned who wrote this benchmark, Is this a significant regression?
Project Member

Comment 14 by 42576172...@developer.gserviceaccount.com, Oct 13 2016


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


===== SUSPECTED CL(s) =====
Subject : Add LevelDB database names to tracing path for easier debugging
Author  : ssid
Commit description:
  
The data from field trials show large databases, but not sure what are
their names. So, adding names to the path.

BUG= 645126 

Review-Url: https://codereview.chromium.org/2399903004
Cr-Commit-Position: refs/heads/master@{#424055}
Commit  : 279f8d702e8b2fffda1b7f50b311c5202331e121
Date    : Sat Oct 08 03:21:14 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N  Good?
chromium@424012  2881.79  39.0689  5  good
chromium@424042  2897.59  19.5091  5  good
chromium@424050  2881.12  64.5898  5  good
chromium@424054  2712.14  78.972   5  good
chromium@424055  2479.13  35.4708  5  bad    <--
chromium@424056  2419.38  14.1561  5  bad
chromium@424057  2459.23  37.5299  5  bad
chromium@424071  2434.11  32.6227  5  bad

Bisect job ran on: android_nexus7_perf_bisect
Bug ID: 655140

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests dromaeo.domcorequery
Test Metric: dom/dom
Relative Change: 15.53%
Score: 99.8

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus7_perf_bisect/builds/3395
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8998895386487162352


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=6417643643338752

| 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!
NextAction: 2016-10-20
I agree that looking at the graph, it's hard to tell whether this is just an outlier due to noise.

Let wait for a few week, look at the graph again & decide.

Comment 16 by ssid@chromium.org, Oct 13 2016

Cc: dskiba@chromium.org
The bisect shows values that look very consistent in comment #14. But, the code that is called only on tracing. I am confused.
ssid@: can you make a speculative revert to see if the benchmark recovered?
Super noisy graph, but it's long term trend seems worse that it was before the regression, although there appears to have been some recovery.  I doubt we can bisect this :/
Project Member

Comment 20 by 42576172...@developer.gserviceaccount.com, Oct 26 2016


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


===== SUSPECTED CL(s) =====
Subject : [Autofill] Log count histogram for the number of server cards
Author  : mathp
Commit description:
  
BUG= 653636 
TEST=AutofillMetrics

Review-Url: https://codereview.chromium.org/2395243002
Cr-Commit-Position: refs/heads/master@{#424053}
Commit  : 2af1bc9bc29520e5b3893b8c31d008d1b384b296
Date    : Sat Oct 08 03:04:41 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@423850  2707.03  70.8167  12  good
chromium@424000  2809.3   79.0863  5   good
chromium@424038  2801.87  83.8838  5   good
chromium@424048  2858.89  66.389   8   good
chromium@424051  2860.53  82.0628  8   good
chromium@424052  2893.07  64.9662  8   good
chromium@424053  2654.0   109.568  5   bad    <--
chromium@424057  2472.92  31.7114  5   bad
chromium@424075  2560.72  62.9558  5   bad
chromium@424150  2553.53  117.656  8   bad

Bisect job ran on: android_nexus7_perf_bisect
Bug ID: 655140

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests dromaeo.domcorequery
Test Metric: dom/dom
Relative Change: 7.43%
Score: 98.0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus7_perf_bisect/builds/3426
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8997725546431567584


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5861809947410432

| 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!
Status: Assigned (was: Untriaged)
Status: WontFix (was: Assigned)
This bug is over 8 months old and bisect pointed to different CLs. I think it's likely noise.

Sign in to add a comment