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

Issue 680512 link

Starred by 2 users
Status: Fixed
Owner:
tkent@chromium.org
Closed: Mar 2017
Cc:
tkent@chromium.org
█ meade@chromium.org
ssid@chromium.org
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

3%-5.9% regression in speedometer at 439378:439416

Project Member Reported by hablich@chromium.org, Jan 12 2017 
See the link to graphs below.
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Jan 12 2017 


=== BISECT JOB RESULTS ===
NO Perf regression found

Bisect Details
  Configuration: win_8_perf_bisect
  Benchmark    : speedometer
  Metric       : jQuery-TodoMVC/jQuery-TodoMVC

Revision             Result                  N
chromium@439389      790.424 +- 126.239      60      good
chromium@439409      796.978 +- 122.486      60      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests speedometer

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8990661465929912800

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5826564580704256


| 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 5 by 42576172...@developer.gserviceaccount.com, Jan 16 2017

Cc: fdoray@chromium.org
Owner: fdoray@chromium.org

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

Hi fdoray@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 : fdoray
  Commit : ad58f8382643800d7107f43c91bd2f9776da4900
  Date   : Mon Dec 19 18:11:30 2016
  Subject: Run ScopedTaskScheduler tasks synchronously.

Bisect Details
  Configuration: win_x64_perf_bisect
  Benchmark    : speedometer
  Metric       : jQuery-TodoMVC/jQuery-TodoMVC
  Change       : 7.12% | 788.387333333 -> 844.510083333

Revision             Result                  N
chromium@439292      788.387 +- 10.9903      6      good
chromium@439404      780.092 +- 13.4245      6      good
chromium@439460      810.829 +- 10.7667      6      good
chromium@439488      816.05 +- 13.7118       6      good
chromium@439503      799.722 +- 21.8679      6      good
chromium@439504      799.326 +- 15.3335      6      good
chromium@439505      822.133 +- 8.1058       6      bad       <--
chromium@439507      830.087 +- 31.2376      6      bad
chromium@439510      842.514 +- 16.2276      6      bad
chromium@439516      844.51 +- 11.8855       6      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests speedometer

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8990281188339397360

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5310839470424064


| 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 6 by hablich@chromium.org, Jan 18 2017

Status: Assigned (was: Untriaged)

Comment 8 by fdoray@chromium.org, Feb 14 2017

Cc: -hablich@chromium.org -fdoray@chromium.org
Owner: hablich@chromium.org
Status: Untriaged (was: Assigned)
base::test::ScopedTaskScheduler is not linked in the binaries that we ship (base/test/BUILD.gn) and thus cannot be responsible for this regression.

It would make more sense to blame a CL in the regression range http://test-results.appspot.com/revision_range?start=439389&end=439409
Project Member

Comment 9 by 42576172...@developer.gserviceaccount.com, Feb 14 2017

Cc: meade@chromium.org
Owner: meade@chromium.org

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

Hi meade@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 : meade
  Commit : 0eebfb8ffb2b8730c2679d41c011f2871b28cdf1
  Date   : Mon Dec 19 04:12:43 2016
  Subject: Remove double constructors that assume degrees from CSSRotation

Bisect Details
  Configuration: win_8_perf_bisect
  Benchmark    : speedometer
  Metric       : jQuery-TodoMVC/jQuery-TodoMVC
  Change       : 3.08% | 789.052416667 -> 812.953722222

Revision             Result                  N
chromium@439388      789.052 +- 13.4983      6       good
chromium@439399      793.18 +- 21.9956       9       good
chromium@439402      786.668 +- 12.6083      6       good
chromium@439403      782.994 +- 12.5409      6       good
chromium@439404      803.752 +- 28.3488      14      bad       <--
chromium@439409      812.954 +- 18.1562      9       bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests speedometer

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8987663990895763968

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=4917964433784832


| 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 Speed>Bisection.  Thank you!

Comment 10 by meade@chromium.org, Feb 14 2017

Owner: hablich@chromium.org
Hmm, unless Speedometer is calling into the Typed OM for some reason, I don't see how changes to the constructor of one of its objects can affect performance.
Project Member

Comment 13 by 42576172...@developer.gserviceaccount.com, Feb 28 2017

Cc: tkent@chromium.org
Owner: tkent@chromium.org

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

Hi tkent@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 : tkent
  Commit : 03e415cffc61351a6b97d069aedbb22edcfc8aec
  Date   : Mon Dec 19 04:32:15 2016
  Subject: Replace ASSERT, ENABLE(ASSERT), and ASSERT_NOT_REACHED in wtf

Bisect Details
  Configuration: win_x64_perf_bisect
  Benchmark    : speedometer
  Metric       : jQuery-TodoMVC/jQuery-TodoMVC
  Change       : 5.06% | 785.571416667 -> 825.3245

Revision             Result                  N
chromium@439392      785.571 +- 8.21328      6      good
chromium@439404      787.122 +- 8.80392      6      good
chromium@439407      783.822 +- 4.6986       6      good
chromium@439408      789.448 +- 11.8504      6      good
chromium@439409      828.9 +- 7.42368        6      bad       <--
chromium@439410      814.824 +- 12.9554      6      bad
chromium@439416      825.325 +- 4.62383      6      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests speedometer

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8986390422620440224

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=4955754534010880


| 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 Speed>Bisection.  Thank you!
Project Member

Comment 14 by 42576172...@developer.gserviceaccount.com, Feb 28 2017

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, please take a look at the
results.


=== BISECT JOB RESULTS ===
Perf regression found with culprit

Suspected Commit
  Author : ssid
  Commit : 4460b9b6393448adc011207ae63380a70ddba369
  Date   : Mon Dec 19 00:56:21 2016
  Subject: [memory-infra] Add support for polling memory totals in MemoryDumpManager

Bisect Details
  Configuration: win_8_perf_bisect
  Benchmark    : speedometer
  Metric       : FlightJS-TodoMVC/FlightJS-TodoMVC
  Change       : 0.58% | 234.123607143 -> 235.427678571

Revision             Result                  N
chromium@439388      234.124 +- 6.2743       14      good
chromium@439389      234.759 +- 1.59688      9       good
chromium@439390      235.509 +- 4.5436       14      bad       <--
chromium@439392      235.154 +- 5.1193       21      bad
chromium@439394      235.239 +- 3.52289      21      bad
chromium@439399      235.237 +- 4.23254      21      bad
chromium@439409      235.428 +- 1.9708       14      bad

To Run This Test
  src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests speedometer

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8986390429922472624

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5281476800675840


| 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 Speed>Bisection.  Thank you!

Comment 15 by ssid@chromium.org, Feb 28 2017

Owner: tkent@chromium.org
Assigning to tkent since the bisect results show better difference in comment #13 and looking at comment #8
My cl changes only tracing.

Comment 16 by tkent@chromium.org, Mar 2 2017

Status: Started (was: Untriaged)
It seems MSVC doesn't like the HashTable.h change in 03e415cffc61351a6b97d069aedbb22edcfc8aec .
Project Member

Comment 17 by bugdroid1@chromium.org, Mar 2 2017 

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/eecb9420efdea1d6c4e882a85e1de607ed317781

commit eecb9420efdea1d6c4e882a85e1de607ed317781
Author: tkent <tkent@chromium.org>
Date: Thu Mar 02 08:08:13 2017

Fix WTF::HashTable performance regression on Windows.

https://codereview.chromium.org/2585673002 introduced function calls for empty
functions. It seems MSVC doesn't optimize them out.

BUG= 680512 

Review-Url: https://codereview.chromium.org/2725953003
Cr-Commit-Position: refs/heads/master@{#454209}

[modify] https://crrev.com/eecb9420efdea1d6c4e882a85e1de607ed317781/third_party/WebKit/Source/wtf/HashTable.h

Comment 18 by tkent@chromium.org, Mar 5 2017

Status: Fixed (was: Started)
It's difficult to observe the progression by #17 because we had a huge regression and progression a few weeks ago.  I believe #17 fixed this issue.

Comment 19 by tkent@chromium.org, Mar 5 2017

Labels: -M-57
►

Sign in to add a comment