New issue
Advanced search Search tips

Issue 625894 link

Starred by 1 user

Issue metadata

Status: WontFix
Owner:
Closed: Dec 2016
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

1%-11.2% regression in v8.infinite_scroll at 403611:403611

Project Member Reported by qyears...@chromium.org, Jul 5 2016

Issue description

See the link to graphs below.
 
No ref build, but there is a consistent increase across different platforms.

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


===== TESTED REVISIONS =====
Revision                       Mean     Std Dev  N  Good?
chromium@403610                6048230  14821.1  5  good
chromium@403610,v8@b1063f7a41  6046546  6794.11  5  good
chromium@403610,v8@6f920d7d59  6042686  3947.26  5  good
chromium@403610,v8@5927deaaf1  6043829  1043.27  5  good
chromium@403610,v8@317dc0578f  6051042  13641.7  5  good
chromium@403610,v8@486d181928  6046075  16449.1  5  good
chromium@403610,v8@de5beb45b8  6044160  881.689  5  good
chromium@403611                6096176  15731.8  5  bad

Bisect job ran on: linux_perf_bisect
Bug ID: 625894

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests top_10_mobile_memory
Test Metric: memory:chrome:all_processes:reported_by_chrome:v8:allocated_objects_size_max/https___mobile.twitter.com_justinbieber?skip_interstitial_true
Relative Change: 0.79%
Score: 0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/linux_perf_bisect/builds/6577
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9007832599226486752


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5882958880178176

| 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!
Project Member

Comment 7 by 42576172...@developer.gserviceaccount.com, Jul 16 2016


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


=== Bisection aborted ===
The bisect was aborted because The metric values for the initial "good" and "bad" revisions do not represent a clear regression.
Please contact the the team (see below) if you believe this is in error.

=== Warnings ===
The following warnings were raised by the bisect job:

 * Bisect failed to reproduce the regression with enough confidence.

===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@403610  8803286  516179   12  good
chromium@403611  7970078  503637   18  bad

Bisect job ran on: android_nexus5X_perf_bisect
Bug ID: 625894

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --also-run-disabled-tests top_10_mobile_memory
Test Metric: memory:chrome:all_processes:reported_by_chrome:v8:allocated_objects_size_max/http___en.m.wikipedia.org_wiki_Science
Relative Change: 11.18%
Score: 0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus5X_perf_bisect/builds/345
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9007038978180887120


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5791307885182976

| 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!
Project Member

Comment 8 by 42576172...@developer.gserviceaccount.com, Jul 16 2016

Cc: ishell@chromium.org
Owner: ishell@chromium.org

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

Hi ishell@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 : [ic] Use UnseededNumberDictionary as a storage for names in TypeFeedbackMetadata.
Author  : ishell
Commit description:
  
The serializer does not support serialization of HashTables in general because
after deserialization it might be necessary to rehash the table.
However the UnseededNumberDictionary does not require rehashing and this CL allows
them to be serialized.

This CL also changes the shape of UnseededNumberDictionary: the details field is
no longer part of the entry since no one needs it.

BUG=chromium:576312,  chromium:623516 

Review-Url: https://codereview.chromium.org/2102073002
Cr-Commit-Position: refs/heads/master@{#37336}
Commit  : 70318619905dd1a10ac0f063da19cebf6d385ba0
Date    : Tue Jun 28 16:16:12 2016


===== TESTED REVISIONS =====
Revision                       Mean     Std Dev  N  Good?
chromium@403610                7416822  6647.9   5  good
chromium@403610,v8@6b63d524c2  7431656  6300.57  5  good
chromium@403610,v8@61c137c811  7431268  5993.36  5  good
chromium@403610,v8@7031861990  7478584  4428.55  5  bad    <--
chromium@403610,v8@efa7095e3e  7484465  9155.49  5  bad
chromium@403610,v8@588e15c034  7486138  3181.55  5  bad
chromium@403610,v8@b1063f7a41  7488534  6977.38  5  bad
chromium@403611                7480523  5463.36  5  bad

Bisect job ran on: win_perf_bisect
Bug ID: 625894

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests top_10_mobile_memory
Test Metric: memory:chrome:all_processes:reported_by_chrome:v8:allocated_objects_size_max/http___search.yahoo.com_search;_ylt_?p_google
Relative Change: 0.86%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/win_perf_bisect/builds/6703
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9007039033249631616


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5812157803921408

| 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!
Project Member

Comment 9 by 42576172...@developer.gserviceaccount.com, Jul 16 2016


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


===== SUSPECTED CL(s) =====
Subject : [ic] Use UnseededNumberDictionary as a storage for names in TypeFeedbackMetadata.
Author  : ishell
Commit description:
  
The serializer does not support serialization of HashTables in general because
after deserialization it might be necessary to rehash the table.
However the UnseededNumberDictionary does not require rehashing and this CL allows
them to be serialized.

This CL also changes the shape of UnseededNumberDictionary: the details field is
no longer part of the entry since no one needs it.

BUG=chromium:576312,  chromium:623516 

Review-Url: https://codereview.chromium.org/2102073002
Cr-Commit-Position: refs/heads/master@{#37336}
Commit  : 70318619905dd1a10ac0f063da19cebf6d385ba0
Date    : Tue Jun 28 16:16:12 2016


===== TESTED REVISIONS =====
Revision                       Mean     Std Dev  N  Good?
chromium@403610                4684076  3709.97  5  good
chromium@403610,v8@6b63d524c2  4695411  6151.19  5  good
chromium@403610,v8@61c137c811  4695162  4351.44  5  good
chromium@403610,v8@7031861990  4731054  3079.17  5  bad    <--
chromium@403610,v8@efa7095e3e  4730736  3562.03  5  bad
chromium@403610,v8@588e15c034  4731628  3128.65  5  bad
chromium@403610,v8@b1063f7a41  4731534  4322.05  5  bad
chromium@403611                4731834  6136.24  5  bad

Bisect job ran on: android_nexus5X_perf_bisect
Bug ID: 625894

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --also-run-disabled-tests top_10_mobile_memory
Test Metric: memory:chrome:all_processes:reported_by_chrome:v8:effective_size_max/https___mobile.twitter.com_justinbieber?skip_interstitial_true
Relative Change: 1.02%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus5X_perf_bisect/builds/344
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9007038987786111904


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5160263322435584

| 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!
The numbers look odd to me, but the bisect bot seems to have high confidence. ishell, any idea what's going on here?
Wait, I was looking at the wrong thing. This change does seem to cause the regression. ishell please investigate :)
Project Member

Comment 12 by bugdroid1@chromium.org, Jul 26 2016

Comment 13 Deleted

Let's see if #12 will help or not.
ishell: I don't see the numbers coming back down from #12
Perf sheriff ping: reminder to follow up on possible performance issues
Project Member

Comment 17 by bugdroid1@chromium.org, Oct 5 2016

The following revision refers to this bug:
  https://chromium.googlesource.com/v8/v8.git/+/c56222c9bd279c8a94697e21ed8ce1e0813c6347

commit c56222c9bd279c8a94697e21ed8ce1e0813c6347
Author: ishell <ishell@chromium.org>
Date: Wed Oct 05 17:55:26 2016

[ic] Avoid feedback metadata names table reallocations.

An attempt to fix memory regression (r38047) caused another regression
because custom capacity chosen for names dictionary implied reallocations
during initialization in some cases.

BUG= chromium:625894 , chromium:632231 

Review-Url: https://codereview.chromium.org/2394873002
Cr-Commit-Position: refs/heads/master@{#40006}

[modify] https://crrev.com/c56222c9bd279c8a94697e21ed8ce1e0813c6347/src/type-feedback-vector.cc

Update: I don't think I'm seeing a recovery in the graphs (https://chromeperf.appspot.com/group_report?bug_id=625894) In the past 4 days.

Is there anything else that might be tried?
We didn't get the new data yet: the latest results correspond to the beginning of September. 
Cc: primiano@chromium.org sullivan@chromium.org
I don't see much data after the beginning of September. What's going on? How can we verify that this is fixed?
We dropped top_10_mobile_memory upstream a while ago (in favour of system_health.memory_mobile) and we are in the process of doing the same downstream.
I am sorry but we can't do analysis on 6 months old bugs.
Realistically, whatever the cause of this regression was it has been taken and is now "a feature" of chrome.
Alright, sounds like this should be marked WontFix; sullivan@, does that sound right?
Status: WontFix (was: Assigned)

Sign in to add a comment