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

Issue 672083 link

Starred by 1 user

Issue metadata

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



Sign in to add a comment

15.9% regression in smoothness.tough_webgl_cases at 435133:435197

Project Member Reported by tdres...@chromium.org, Dec 7 2016

Issue description

See the link to graphs below.
 
All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=672083

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgl93kvgoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgl5nLpQkM


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

chromium-rel-win7-gpu-ati
Cc: hywu@google.com
Owner: hywu@google.com

=== PERF REGRESSION ===


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

Hi hywu@google.com, the bisect results pointed to your CL, please take a look at the
results.


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


===== SUSPECTED CL(s) =====
Subject : VEA unittest: add a test case for cache line-alignment
Author  : hywu
Commit description:
  
Linux refuses to accept buffers that have sizes ualigned to
CPU cache line, which can be 128 bytes on certain devices.
This test inputs a frame that has unaligned size of U and V
planes to catch such errors.

BUG= 664039 
TEST=Run VEA unittest on Kevin.
CQ_INCLUDE_TRYBOTS=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

Review-Url: https://codereview.chromium.org/2525563005
Cr-Commit-Position: refs/heads/master@{#435179}
Commit  : 62f3a0edf9c5e19da3e59a6b2f5c402895b1805a
Date    : Wed Nov 30 08:53:02 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@435132  62.965   4.1086   12  good
chromium@435165  65.2458  9.49981  12  good
chromium@435173  64.3956  8.90698  12  good
chromium@435177  65.4983  16.6065  18  good
chromium@435178  67.7484  3.72376  8   good
chromium@435179  70.5408  3.80374  8   bad    <--
chromium@435181  69.8172  9.07119  12  bad
chromium@435197  71.0728  12.1659  12  bad

Bisect job ran on: winx64ati_perf_bisect
Bug ID: 672083

Test Command: src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=http...kenrussell.github.io.webgl.animometer.Animometer.tests.3d.webgl.html smoothness.tough_webgl_cases
Test Metric: mean_frame_time/http___kenrussell.github.io_webgl-animometer_Animometer_tests_3d_webgl.html
Relative Change: 11.68%

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/winx64ati_perf_bisect/builds/1658
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8993921707662956864


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5691064617271296

| 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 hywu@chromium.org, Dec 8 2016

The CL modifies only a unit test that runs in Chrome OS only. It is impossible to impose any impact on performance.
Cc: dtu@chromium.org simonhatch@chromium.org
Owner: ----
+simonhatch, dtu: bisect shows a clear shift in mean at a CL which can't possibly be related in #3. Any ideas? On the graphs, the test looks a bit bimodal.

===== BISECT JOB RESULTS =====
Status: failed


=== Bisection aborted ===
The bisect was aborted because Bisect cannot identify a culprit: Bisect failed to reproduce the regression with enough confidence.
Please contact the the team (see below) if you believe this is in error.

===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N    Good?
chromium@435132  69.4831  338.192  358  good
chromium@435197  68.8665  336.046  359  bad

Bisect job ran on: winx64ati_perf_bisect
Bug ID: 672083

Test Command: src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=http...kenrussell.github.io.webgl.animometer.Animometer.tests.3d.webgl.html smoothness.tough_webgl_cases
Test Metric: frame_times/http___kenrussell.github.io_webgl-animometer_Animometer_tests_3d_webgl.html
Relative Change: 0.89%

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/winx64ati_perf_bisect/builds/1661
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8993837882979351504


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5851145799467008

| 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!
Cc: alexande...@intel.com
Owner: alexande...@intel.com

=== PERF REGRESSION ===


=== Auto-CCing suspected CL author alexander.shalamov@intel.com ===

Hi alexander.shalamov@intel.com, the bisect results pointed to your CL, please take a look at the
results.


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


===== SUSPECTED CL(s) =====
Subject : [webnfc] Implement watch method for Android nfc mojo service.
Author  : alexander.shalamov
Commit description:
  
Implementation of nfc.watch [1] and nfc.cancelWatch [2] methods
for Android nfc mojo service that is used from W3C WebNFC module.

[1] https://w3c.github.io/web-nfc/#dom-nfc-watch
[2] https://w3c.github.io/web-nfc/#dom-nfc-cancelwatch

BUG=520391
BUG= 625589 

Review-Url: https://codereview.chromium.org/1765533004
Cr-Commit-Position: refs/heads/master@{#435189}
Commit  : 55ac6a4ca3780a5c315a7014f43642e43c5245f0
Date    : Wed Nov 30 09:07:43 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@435132  72.9811  2.70753  8   good
chromium@435165  70.8609  4.34444  8   good
chromium@435181  70.0291  6.92601  12  good
chromium@435185  72.5564  0.81992  5   good
chromium@435187  70.6554  4.90353  5   good
chromium@435188  70.7608  3.76549  8   good
chromium@435189  66.1605  9.59974  12  bad    <--
chromium@435197  64.0387  6.31317  8   bad

Bisect job ran on: winx64ati_perf_bisect
Bug ID: 672083

Test Command: src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=http...kenrussell.github.io.webgl.animometer.Animometer.tests.3d.webgl.html smoothness.tough_webgl_cases
Test Metric: mean_frame_time/http___kenrussell.github.io_webgl-animometer_Animometer_tests_3d_webgl.html
Relative Change: 10.94%

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/winx64ati_perf_bisect/builds/1660
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8993838214531715824


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5894672071786496

| 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 11 by 42576172...@developer.gserviceaccount.com, Dec 12 2016


===== BISECT JOB RESULTS =====
Status: failed


=== Bisection aborted ===
The bisect was aborted because Bisect cannot identify a culprit: Bisect failed to reproduce the regression with enough confidence.
Please contact the the team (see below) if you believe this is in error.

===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@435132  68.0068  25.0358  21  good
chromium@435186  66.6826  24.2946  21  bad

Bisect job ran on: winx64ati_perf_bisect
Bug ID: 672083

Test Command: src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=http...kenrussell.github.io.webgl.animometer.Animometer.tests.3d.webgl.html smoothness.tough_webgl_cases
Test Metric: mean_frame_time/http___kenrussell.github.io_webgl-animometer_Animometer_tests_3d_webgl.html
Relative Change: 1.95%

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/winx64ati_perf_bisect/builds/1664
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8993493664917284176


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5846868079149056

| 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 13 by 42576172...@developer.gserviceaccount.com, Dec 12 2016


===== BISECT JOB RESULTS =====
Status: failed


=== Bisection aborted ===
The bisect was aborted because Bisect cannot identify a culprit: Bisect failed to reproduce the regression with enough confidence.
Please contact the the team (see below) if you believe this is in error.

===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@435188  67.6491  9.1626   21  good
chromium@435189  69.6914  13.2632  21  bad

Bisect job ran on: winx64ati_perf_bisect
Bug ID: 672083

Test Command: src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=http...kenrussell.github.io.webgl.animometer.Animometer.tests.3d.webgl.html smoothness.tough_webgl_cases
Test Metric: mean_frame_time/http___kenrussell.github.io_webgl-animometer_Animometer_tests_3d_webgl.html
Relative Change: 3.02%

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/winx64ati_perf_bisect/builds/1665
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8993485819842287056


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5876432889184256

| 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!
CL chromium@435189 adds Android specific implementation that should not affect performance of WebGL related tests, especially on win bots. Moreover, blink side code is under experimental flag, thus, cannot be accessed by perf tests at runtime. In 435133 - 435197 commit range, there are few WebGL / Skia related commits, could those somehow affect performance?
Owner: ----
simonhatch, dtu, can you take a look at this? Both #3 and #9 found a clear shift in the median on separate CLs that couldn't possibly be related.
This is really weird because #3 and #9 reproduce the same relative change, but reversed? Very bizarre, only difference I see is they're different bots (#3: build104-m4, #9: build103-m4).


Agree with #c7 that the test looked bimodal pre-regression, and the regression seems to have forced it into always getting the upper value. It shows up on the other win gpu bots too, but as the point when it goes from relatively stable to noisy. I tried kicking off a bisect on the intel one.
Project Member

Comment 19 by 42576172...@developer.gserviceaccount.com, Dec 13 2016


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


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@434582  74.3535  15.2891  21  good
chromium@435386  72.7608  1.61401  6   good
chromium@435782  73.2567  1.95441  9   good
chromium@435987  75.8298  17.5232  14  good
chromium@436087  73.515   1.44464  14  good
chromium@436088  73.7813  3.11248  21  unknown
chromium@436089  73.8742  3.2384   21  unknown
chromium@436090  74.8387  4.99649  21  unknown
chromium@436091  73.7879  2.46314  21  unknown
chromium@436116  73.8103  2.5295   21  unknown
chromium@436117  73.5188  3.49409  14  good
chromium@436123  73.64    2.56676  14  good
chromium@436126  73.706   2.9316   21  good
chromium@436127  76.1938  19.4525  21  unknown
chromium@436128  73.5844  2.18258  21  unknown
chromium@436129  73.7042  3.42691  21  unknown
chromium@436130  73.7075  3.08557  21  unknown
chromium@436131  74.893   6.39982  14  bad
chromium@436141  74.0839  2.23162  14  bad
chromium@436191  74.526   9.50956  21  bad

Bisect job ran on: winx64intel_perf_bisect
Bug ID: 672083

Test Command: src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=http...kenrussell.github.io.webgl.animometer.Animometer.tests.3d.webgl.html smoothness.tough_webgl_cases
Test Metric: mean_frame_time/http___kenrussell.github.io_webgl-animometer_Animometer_tests_3d_webgl.html
Relative Change: 0.23%

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/winx64intel_perf_bisect/builds/1252
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8993459862142880016


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5339991205478400

| 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 20 by 42576172...@developer.gserviceaccount.com, Dec 13 2016

Cc: mlamouri@chromium.org
Owner: mlamouri@chromium.org

=== PERF REGRESSION ===


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

Hi mlamouri@chromium.org, the bisect results pointed to your CL, please take a look at the
results.


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


===== SUSPECTED CL(s) =====
Subject : Media Controls: handle 'timeupdate', 'play' and 'pause' via an EventListener.
Author  : mlamouri
Commit description:
  
Instead of having hard coded plumbing from HTMLMediaElement to MediaControls.

BUG= 662761 

Review-Url: https://codereview.chromium.org/2499883002
Cr-Commit-Position: refs/heads/master@{#435220}
Commit  : 88ea177babd78b83478c082a903c45e8b3499703
Date    : Wed Nov 30 11:41:45 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev   N   Good?
chromium@435083  60.678   0.785707  6   good
chromium@435219  61.2921  3.4274    14  good
chromium@435220  67.742   7.92039   6   bad    <--
chromium@435221  65.7062  10.1382   9   bad
chromium@435222  65.5182  5.36499   6   bad
chromium@435224  65.1063  5.45297   6   bad
chromium@435228  65.5965  16.575    21  bad
chromium@435236  65.3729  8.91073   14  bad
chromium@435253  65.833   9.20402   9   bad
chromium@435287  65.7342  13.3834   14  bad
chromium@435354  69.9307  4.56845   9   bad
chromium@435624  71.436   1.62374   6   bad
chromium@436164  72.8672  0.813899  6   bad

Bisect job ran on: winx64ati_perf_bisect
Bug ID: 672083

Test Command: src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests smoothness.tough_webgl_cases
Test Metric: mean_frame_time/http___kenrussell.github.io_webgl-animometer_Animometer_tests_3d_webgl.html
Relative Change: 20.09%

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/winx64ati_perf_bisect/builds/1668
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8993463201695160368


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5824507808317440

| 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: ----
Unassigning myself as I doubt it has anything to do with my change that is abut media element controls and the events the buttons use to change state.
Project Member

Comment 23 by 42576172...@developer.gserviceaccount.com, Dec 13 2016

Cc: mkwst@chromium.org
Owner: mkwst@chromium.org

=== PERF REGRESSION ===


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

Hi mkwst@chromium.org, the bisect results pointed to your CL, please take a look at the
results.


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


===== SUSPECTED CL(s) =====
Subject : CSP: "local schemes" should inherit policy when embedded.
Author  : mkwst
Commit description:
  
https://w3c.github.io/webappsec-csp/#initialize-document-csp mandates
that resources with "local schemes" ('data:', 'blob:', 'filesystem:',
'about:') inherit the policy of their embedding context when pulled in
via an '<iframe>'.

I'm pretty sure this worked at some point in the past, but I apparently
didn't put a test on it. It does work in Firefox. Let's match their
behavior and lock it in.

BUG= 513860 
R=jochen@chromium.org, dcheng@chromium.org

Review-Url: https://codereview.chromium.org/2472333003
Cr-Commit-Position: refs/heads/master@{#435165}
Commit  : e82aced182b4fe7ac99437943a7c39238796c1d9
Date    : Wed Nov 30 08:28:05 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@435132  62.2918  7.06108  21  good
chromium@435149  63.202   5.51186  6   good
chromium@435157  61.9663  4.5148   6   good
chromium@435161  68.5257  7.67219  6   good
chromium@435163  68.881   6.75612  6   good
chromium@435164  62.1925  3.12844  6   good
chromium@435165  72.8882  1.1583   6   bad    <--
chromium@435197  65.8002  18.5721  21  bad

Bisect job ran on: winx64ati_perf_bisect
Bug ID: 672083

Test Command: src/tools/perf/run_benchmark -v --browser=release_x64 --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=http...kenrussell.github.io.webgl.animometer.Animometer.tests.3d.webgl.html smoothness.tough_webgl_cases
Test Metric: mean_frame_time/http___kenrussell.github.io_webgl-animometer_Animometer_tests_3d_webgl.html
Relative Change: 5.63%

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/winx64ati_perf_bisect/builds/1669
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8993383127745927072


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5808770108620800

| 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)
Owner: ----
Status: WontFix (was: Assigned)
Doesn't look like we'll get to the root cause of this one.

Sign in to add a comment