New issue
Advanced search Search tips

Issue 647635 link

Starred by 1 user

Issue metadata

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



Sign in to add a comment

3.2%-6.3% regression in speedometer at 418870:418924

Project Member Reported by nikolaos@chromium.org, Sep 16 2016

Issue description

See the link to graphs below.
 
All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=647635

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg2ayYlwgM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg2fid4wgM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgue7_sQoM


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

chromium-rel-mac11
chromium-rel-win7-x64-dual
chromium-rel-win8-dual
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Sep 16 2016

Cc: agrieve@chromium.org
Owner: agrieve@chromium.org

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

Hi agrieve@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 : Move repack_pseudo_locales //chrome -> //chrome/test
Author  : agrieve
Commit description:
  
According to https://www.chromium.org/developers/testing/fake-bidi, it
is no longer usable via --locale_pak, and so is used only by the
(disabled) test bidi_checker_web_ui_test.cc.

BUG=645716

Review-Url: https://codereview.chromium.org/2339513002
Cr-Commit-Position: refs/heads/master@{#418889}
Commit  : cddda4b3eebaacab696a28aa18f89f92aceda3a9
Date    : Thu Sep 15 17:10:08 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N  Good?
chromium@418877  757.978  2.14984  5  good
chromium@418885  779.493  6.41567  5  good
chromium@418887  782.193  11.0235  5  good
chromium@418888  786.719  7.41706  5  good
chromium@418889  812.385  5.6273   5  bad    <--
chromium@418893  801.287  1.02028  5  bad

Bisect job ran on: win_x64_perf_bisect
Bug ID: 647635

Test Command: src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --also-run-disabled-tests speedometer
Test Metric: jQuery-TodoMVC/jQuery-TodoMVC
Relative Change: 5.71%
Score: 99.5

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/win_x64_perf_bisect/builds/1472
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9001369782443401104


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5824175288090624

| 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: nikolaos@chromium.org
Change was a test-only refactoring. Couldn't have affected the build afaict.
Status: WontFix (was: Assigned)
Cc: cbruni@chromium.org verwa...@chromium.org hablich@chromium.org
Status: Assigned (was: WontFix)
I just did a manual comparison run between Chrome 55 Canary and Chrome 54 Beta on Speedometer. Scores are runs per minute (higher is better):

55 canary:
117.1
117.6
118.4

54 beta:
127.5
126.9
127.1

There was a regression happening on ToT => reopening the bug.
Kicking off a few more bisects with clobbered builds. Let's see.
Project Member

Comment 11 by 42576172...@developer.gserviceaccount.com, Sep 21 2016

Cc: hirosh...@chromium.org
Owner: hirosh...@chromium.org

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

Hi hiroshige@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 : Make internals.isPreloaded() to remain the same before/after clearPreloads()
Author  : hiroshige
Commit description:
  
Previously, internals.isPreloaded() depends on |ResourceFetcher::m_preloads|,
which is cleared by ResourceFetcher::clearPreloads().
This caused internals.isPreloaded() to turn false after around document's
load event.

This CL makes ResourceFetcher to keep a set of preloaded URLs (separately
from |m_preloads|) if a blink::Internals object is created, and use the set
to calculate internals.isPreloaded().

BUG= 643621 

Review-Url: https://codereview.chromium.org/2332333003
Cr-Commit-Position: refs/heads/master@{#418821}
Commit  : 7978814bba57818814b7a10aa3366e16c6d33784
Date    : Thu Sep 15 10:07:33 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@418769  1258.83  21.6203  12  good
chromium@418798  1260.22  9.67084  5   good
chromium@418813  1239.73  14.0796  8   good
chromium@418820  1248.71  15.1815  18  good
chromium@418821  1262.04  16.2887  18  bad    <--
chromium@418822  1274.9   20.4072  12  bad
chromium@418824  1274.35  8.08934  8   bad
chromium@418827  1280.07  14.2913  12  bad
chromium@418884  1287.63  11.7437  12  bad
chromium@418999  1316.1   26.0461  8   bad

Bisect job ran on: mac_10_11_perf_bisect
Bug ID: 647635

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests speedometer
Test Metric: React-TodoMVC/React-TodoMVC
Relative Change: 3.81%
Score: 99.8

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/mac_10_11_perf_bisect/builds/916
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9000908280678514464


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5051174389022720

| 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!
> C#11

My change is for layout test and shouldn't affect the behavior unless blink::Internals is created.

Also, the bisect result doesn't look significant:

chromium@418798  1260.22  9.67084  5   good   <--
chromium@418821  1262.04  16.2887  18  bad    <--

Re-running bisects.
Project Member

Comment 15 by 42576172...@developer.gserviceaccount.com, Sep 21 2016

Cc: bsalo...@google.com
Owner: bsalo...@google.com

=== Auto-CCing suspected CL author bsalomon@google.com ===

Hi bsalomon@google.com, 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 : Remove "f" suffix from literal in LCD DF text shader.
Author  : bsalomon
Commit description:
  
Also improve DF GP names
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2340603003

Review-Url: https://codereview.chromium.org/2340603003
Commit  : 32d1e95ca593adfaa12bc48d4ac0d5501a4b2046
Date    : Thu Sep 15 14:29:52 2016


===== TESTED REVISIONS =====
Revision                         Mean     Std Dev   N  Good?
chromium@418177                  783.383  0.916814  5  good
chromium@418581                  757.488  5.87542   5  good
chromium@418797                  759.949  5.62115   5  good
chromium@418848                  760.591  3.70479   5  good
chromium@418873                  774.28   5.68569   5  good
chromium@418880                  762.264  2.00222   5  good
chromium@418883                  770.927  2.7974    5  good
chromium@418885                  768.522  7.09481   8  good
chromium@418885,skia@1dfec74d6b  773.079  8.84836   8  good
chromium@418885,skia@32d1e95ca5  797.875  7.67219   5  bad    <--
chromium@418885,skia@e839e78443  787.775  2.31396   8  bad
chromium@418886                  789.942  5.47406   8  bad
chromium@418898                  799.241  4.61361   5  bad
chromium@418999                  807.777  3.91784   5  bad

Bisect job ran on: win_x64_perf_bisect
Bug ID: 647635

Test Command: src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --also-run-disabled-tests speedometer
Test Metric: jQuery-TodoMVC/jQuery-TodoMVC
Relative Change: 3.11%
Score: 99.5

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/win_x64_perf_bisect/builds/1477
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9000908305634125424


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=6465025814822912

| 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 16 by 42576172...@developer.gserviceaccount.com, Sep 21 2016

Cc: vabr@chromium.org
Owner: vabr@chromium.org

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

Hi vabr@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 : c/browser, c/common, components O-P: Change ScopedTempDir::path() to GetPath()
Author  : vabr
Commit description:
  
path() is being deprecated, GetPath() has better checking against wrong use.

For more context, see https://codereview.chromium.org/2275553005/.

BUG= 640599 

Review-Url: https://codereview.chromium.org/2317123002
Cr-Commit-Position: refs/heads/master@{#418804}
Commit  : 8023d87a36c3eb4aacbf3e9ea427ecf9b572880f
Date    : Thu Sep 15 08:14:14 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@418777  233.491  2.80836  41  good
chromium@418791  234.161  2.12538  41  good
chromium@418798  233.261  2.32761  62  good
chromium@418802  234.22   2.01883  62  good
chromium@418803  234.9    2.47903  41  good
chromium@418804  236.284  2.30972  41  bad    <--
chromium@418805  235.588  2.31297  41  bad
chromium@418833  236.881  2.27075  18  bad
chromium@418888  237.61   3.17779  12  bad
chromium@418999  238.604  2.97788  12  bad

Bisect job ran on: win_8_perf_bisect
Bug ID: 647635

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests speedometer
Test Metric: VanillaJS-TodoMVC/VanillaJS-TodoMVC
Relative Change: 1.40%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/win_8_perf_bisect/builds/2202
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9000908258686546432


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=6360714506141696

| 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 17 by 42576172...@developer.gserviceaccount.com, Sep 22 2016

Cc: v8-autoroll@chromium.org
Owner: v8-autoroll@chromium.org

=== Auto-CCing suspected CL author v8-autoroll@chromium.org ===

Hi v8-autoroll@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 : Update V8 to version 5.5.208.
Author  : v8-autoroll
Commit description:
  
Summary of changes available at:
https://chromium.googlesource.com/v8/v8/+log/6f1ef12e..fa07ac00

Please follow these instructions for assigning/CC'ing issues:
https://github.com/v8/v8/wiki/Triaging%20issues

Please close rolling in case of a roll revert:
https://v8-roll.appspot.com/
This only works with a Google account.

CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_precise_blink_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel

TBR=hablich@chromium.org,machenbach@chromium.org,littledan@chromium.org,vogelheim@chromium.org

Review-Url: https://codereview.chromium.org/2347653002
Cr-Commit-Position: refs/heads/master@{#418891}
Commit  : c08e4b2781c2f42ef7b55604aa107d46e3abb265
Date    : Thu Sep 15 17:15:09 2016


===== TESTED REVISIONS =====
Revision                       Mean     Std Dev  N   Good?
chromium@418869                1281.09  15.2686  12  good
chromium@418881                1284.84  12.6789  27  good
chromium@418887                1289.03  19.154   27  good
chromium@418890                1293.23  19.3298  27  good
chromium@418890,v8@b848716c98  1309.08  12.3845  41  good
chromium@418890,v8@0e938d4ac0  1310.31  15.2546  41  good
chromium@418890,v8@fa07ac0055  1296.89  11.0483  5   good
chromium@418891                1320.02  16.5563  27  bad    <--
chromium@418892                1308.64  16.9924  8   bad
chromium@418893                1305.66  15.3097  27  bad

Bisect job ran on: mac_10_11_perf_bisect
Bug ID: 647635

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests speedometer
Test Metric: React-TodoMVC/React-TodoMVC
Relative Change: 2.65%
Score: 98.0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/mac_10_11_perf_bisect/builds/917
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9000885430788857840


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=6097221995462656

| 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 18 by 42576172...@developer.gserviceaccount.com, Sep 22 2016

Cc: a...@chromium.org
Owner: a...@chromium.org

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

Hi avi@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 : Remove QUIC's Interval::difference(vector<Interval*>*).
Author  : avi
Commit description:
  
It's unused.

BUG= 555865 
R=bnc@chromium.org

Review-Url: https://codereview.chromium.org/2339223003
Cr-Commit-Position: refs/heads/master@{#418890}
Commit  : 04d50002836c9cc18d131e450d1db50d5cf67cd5
Date    : Thu Sep 15 17:12:55 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N    Good?
chromium@418884  1279.78  17.2301  27   good
chromium@418888  1289.92  15.2446  140  good
chromium@418889  1289.23  17.3783  140  good
chromium@418890  1295.29  15.8478  93   bad    <--
chromium@418892  1299.53  12.0775  12   bad
chromium@418899  1293.21  13.1079  18   bad
chromium@418913  1291.99  15.8509  27   bad
chromium@418942  1302.22  18.3979  27   bad
chromium@418999  1300.45  12.0546  18   bad

Bisect job ran on: mac_10_11_perf_bisect
Bug ID: 647635

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests speedometer
Test Metric: React-TodoMVC/React-TodoMVC
Relative Change: 0.63%
Score: 98.0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/mac_10_11_perf_bisect/builds/918
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9000885414096035616


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=6696679707770880

| 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 19 by a...@chromium.org, Sep 22 2016

Owner: hablich@chromium.org
This can't be that CL. That CL removed dead code, literally called nowhere.

Filed  bug 649352  about the AutoBisect spamming.
FWIW the Skia CL that was cited in #15 also is extremely unlikely to be at fault. It changed a literal from "1.0f" to "1.0" in a GLSL shader.
I is strange that those alerts cannot be reproduced. I rechecked 55 vs 54 again with --enable-benchmarking. The performance is now approx. the same. My assumption is that there was some hiccup with that flag recently and Ignition was running? Because when I run the benchmarks with Ignition explicitly switched on, I get a similar tankage.
One of the issues is that the metrics being bisected on are quite noisy. You can tell that there was a regression when you look at the average between runs before and after the revisions range, but not necessarily by picking individual revisions on either side of it. VanillaJS-TodoMVC in particular isn't a useful metric to bisect on in this case.

Comment 23 by vabr@chromium.org, Sep 27 2016

Cc: -vabr@chromium.org
(Seems like I have nothing to contribute here, feel free to re-add me if you disagree.)
Status: WontFix (was: Assigned)
Closing this ... seems to be flaky bots.

Sign in to add a comment