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

Issue 696850 link

Starred by 1 user
Status: Fixed
Owner:
samans@chromium.org
Closed: Mar 2017
Cc:
█ staraz@chromium.org
samans@chromium.org
█ fsam...@chromium.org
tguilbert@chromium.org
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression

Blocking:
issue 696881



Sign in to add a comment

8%-44.2% regression in media.tough_video_cases at 452898:453062 [memory]

Project Member Reported by tguilbert@chromium.org, Feb 28 2017 
See the link to graphs below.

Comment 3 by tguilbert@chromium.org, Feb 28 2017

Summary: 8%-44.2% regression in media.tough_video_cases at 452898:453062 [memory] (was: 8%-44.2% regression in media.tough_video_cases at 452898:453062)

Comment 4 by tguilbert@chromium.org, Feb 28 2017

Labels: media_regression_2_27

Comment 5 by tguilbert@chromium.org, Feb 28 2017

Blocking: 696881
Project Member

Comment 6 by 42576172...@developer.gserviceaccount.com, Feb 28 2017

Cc: samans@chromium.org
Owner: samans@chromium.org

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

Hi samans@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 : samans
  Commit : 566aae768dfb729fad06c14339dcd7dfe2e5aba5
  Date   : Fri Feb 24 23:39:30 2017
  Subject: DelegatedFrameHost should use CompositorFrameSinkSupport

Bisect Details
  Configuration: win_x64_perf_bisect
  Benchmark    : media.tough_video_cases
  Metric       : vm_working_set_delta_size/smpte_3840x2160_60fps_vp9.webm_browser
  Change       : 59.56% | 15869.3333333 -> 25320.6666667

Revision             Result                  N
chromium@452920      15869.3 +- 1915.11      6      good
chromium@452989      16117.3 +- 2030.51      6      good
chromium@452998      16425.3 +- 1197.94      6      good
chromium@452999      23223.3 +- 1270.96      6      bad       <--
chromium@453000      23013.3 +- 2215.19      6      bad
chromium@453001      26562.7 +- 13554.9      6      bad
chromium@453003      24938.7 +- 10029.3      6      bad
chromium@453007      23061.3 +- 1577.95      6      bad
chromium@453024      22734.0 +- 1277.62      6      bad
chromium@453058      25320.7 +- 14112.4      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 media.tough_video_cases

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8986451616224408848

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5871514374635520


| 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 7 by samans@chromium.org, Feb 28 2017

Cc: staraz@chromium.org

Comment 8 by samans@chromium.org, Feb 28 2017

Cc: fsam...@chromium.org

Comment 9 by samans@chromium.org, Mar 3 2017

Owner: tguilbert@chromium.org

Comment 15 by tguilbert@chromium.org, Mar 3 2017 

I have started new bisects, on a wider range of tests.
Project Member

Comment 16 by 42576172...@developer.gserviceaccount.com, Mar 3 2017 


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

Suspected Commit
  Author : samans
  Commit : 566aae768dfb729fad06c14339dcd7dfe2e5aba5
  Date   : Fri Feb 24 23:39:30 2017
  Subject: DelegatedFrameHost should use CompositorFrameSinkSupport

Bisect Details
  Configuration: winx64_10_perf_bisect
  Benchmark    : media.tough_video_cases
  Metric       : vm_working_set_delta_size/video.html?src_crowd1080.ogv
  Change       : 12.67% | 97698.6666667 -> 110074.666667

Revision             Result                  N
chromium@452897      97698.7 +- 2383.47      6      good
chromium@452954      96946.7 +- 3637.62      6      good
chromium@452982      97734.0 +- 6115.61      6      good
chromium@452996      97680.7 +- 1490.05      6      good
chromium@452998      98312.0 +- 5813.09      6      good
chromium@452999      110234 +- 3788.93       6      bad       <--
chromium@453000      111337 +- 12321.6       6      bad
chromium@453003      110880 +- 989.109       6      bad
chromium@453010      110075 +- 2463.03       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 --story-filter=video.html.src.crowd1080.ogv media.tough_video_cases

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8986120691566509168

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5815156686192640


| 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 17 by 42576172...@developer.gserviceaccount.com, Mar 3 2017 


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

Suspected Commit
  Author : samans
  Commit : 566aae768dfb729fad06c14339dcd7dfe2e5aba5
  Date   : Fri Feb 24 23:39:30 2017
  Subject: DelegatedFrameHost should use CompositorFrameSinkSupport

Bisect Details
  Configuration: winx64_10_perf_bisect
  Benchmark    : media.tough_video_cases
  Metric       : vm_working_set_delta_size/video.html?src_crowd1080.mp4
  Change       : 8.20% | 114594.285714 -> 123989.714286

Revision             Result                 N
chromium@452897      114594 +- 6956.66      14      good
chromium@452954      114537 +- 3534.15      9       good
chromium@452982      115062 +- 11152.5      9       good
chromium@452996      114683 +- 2598.21      9       good
chromium@452998      113583 +- 4778.46      6       good
chromium@452999      124759 +- 15836.4      9       bad       <--
chromium@453000      127291 +- 1686.31      6       bad
chromium@453003      123626 +- 21215.7      14      bad
chromium@453010      123990 +- 24035.1      14      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 --story-filter=video.html.src.crowd1080.mp4 media.tough_video_cases

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8986120639877574480

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5881207646060544


| 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 18 by tguilbert@chromium.org, Mar 3 2017

Owner: samans@chromium.org
It seems like this is a valid memory regression on windows.

samans@, can you take a look?
Project Member

Comment 19 by 42576172...@developer.gserviceaccount.com, Mar 4 2017 


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

Bisect Details
  Configuration: win_perf_bisect
  Benchmark    : media.tough_video_cases
  Metric       : vm_working_set_delta_size/smpte_3840x2160_60fps_vp9.webm_renderer

Revision             Result                 N
chromium@452930      247794 +- 854.571      21      good
chromium@453062      249278 +- 32025.3      21      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 media.tough_video_cases

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8986120571511887408

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5249377691500544


| 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 20 by 42576172...@developer.gserviceaccount.com, Mar 4 2017 


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

Bisect Details
  Configuration: win_perf_bisect
  Benchmark    : media.tough_video_cases
  Metric       : vm_working_set_delta_size/crowd2160.webm_browser

Revision             Result                  N
chromium@452930      27180.2 +- 1610.28      21      good
chromium@453062      28980.8 +- 20244.4      21      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 media.tough_video_cases

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8986120612551155248

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5852268793954304


| 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 21 by 42576172...@developer.gserviceaccount.com, Mar 4 2017 


=== BISECT JOB RESULTS ===
Perf regression found but unable to narrow commit range

Build failures prevented the bisect from narrowing the range further.


Bisect Details
  Configuration: winx64_10_perf_bisect
  Benchmark    : media.tough_video_cases
  Metric       : vm_working_set_delta_size/crowd1080.mp4_renderer
  Change       : 5.57% | 100528.285714 -> 106131.142857

Suspected Commit Range
  2 commits in range
  https://chromium.googlesource.com/chromium/src/+log/37da9b3e656ac85c42e3f375e84b55c98b397da1..566aae768dfb729fad06c14339dcd7dfe2e5aba5


Revision             Result                 N
chromium@452897      100528 +- 766.341      14       good
chromium@452954      100224 +- 3719.42      9        good
chromium@452982      100471 +- 351.581      9        good
chromium@452996      100413 +- 2154.0       14       good
chromium@452997      100491 +- 1539.6       14       good
chromium@452998      ---                    ---      build failure
chromium@452999      104388 +- 18437.6      21       bad
chromium@453000      105206 +- 14867.1      14       bad
chromium@453003      107786 +- 7806.01      9        bad
chromium@453010      106131 +- 15501.2      14       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 media.tough_video_cases

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8986120532750957232

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5497991604994048


| 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 24 by 42576172...@developer.gserviceaccount.com, Mar 7 2017 


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

Bisect Details
  Configuration: winx64_10_perf_bisect
  Benchmark    : media.tough_video_cases
  Metric       : vm_working_set_delta_size/crowd1080.mp4_renderer

Revision             Result                 N
chromium@452897      100622 +- 1146.7       21      good
chromium@452998      100625 +- 617.616      21      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 media.tough_video_cases

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8985830520025234000

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5848072879341568


| 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 26 by 42576172...@developer.gserviceaccount.com, Mar 7 2017 


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

Bisect Details
  Configuration: winx64_10_perf_bisect
  Benchmark    : media.tough_video_cases
  Metric       : vm_working_set_delta_size/crowd1080.mp4_renderer

Revision             Result                 N
chromium@451853      100242 +- 4109.02      21      good
chromium@452845      100623 +- 1201.17      21      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 media.tough_video_cases

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8985828948908217792

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5838576337551360


| 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 27 by bugdroid1@chromium.org, Mar 9 2017 

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/f0f14b715fc62c3e8d9888970ebda21ea7a6e399

commit f0f14b715fc62c3e8d9888970ebda21ea7a6e399
Author: samans <samans@chromium.org>
Date: Thu Mar 09 02:38:03 2017

CompositorFrameSinkSupport should return resources before sending an ack

Apparently after we switched to CompositorFrameSinkSupport in
DelegatedFrameHost, we started using more memory in some performance tests.
We hypothesized that this could be because we send an ack back to the
client before returning the resources and therefore the client cannot reuse
the old resources and has to allocate new ones. I ran the performance tests
that regressed on this CL and I can see clear improvements in memory usage.

Examples of improved metrics (from win_x64_perf_bisect Build #1626):
vm_working_set_delta_size.smpte_3840x2160_60fps_vp9.webm_total from 297.5 MiB to 258.6 MiB
vm_working_set_delta_size.crowd2160.webm_total from 301.1 MiB to 261.3 MiB

BUG= 696850 
CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_trusty_blink_rel

Review-Url: https://codereview.chromium.org/2734783006
Cr-Commit-Position: refs/heads/master@{#455646}

[modify] https://crrev.com/f0f14b715fc62c3e8d9888970ebda21ea7a6e399/cc/surfaces/compositor_frame_sink_support.cc
[modify] https://crrev.com/f0f14b715fc62c3e8d9888970ebda21ea7a6e399/cc/surfaces/compositor_frame_sink_support_unittest.cc

Comment 28 by samans@chromium.org, Mar 9 2017

Status: Fixed (was: Untriaged)
All the graphs https://chromeperf.appspot.com/group_report?bug_id=696850 are back to normal. I'm marking this as fixed.
Project Member

Comment 30 by 42576172...@developer.gserviceaccount.com, Apr 12 2017 


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

Bisect Details
  Configuration: win_perf_bisect
  Benchmark    : media.tough_video_cases
  Metric       : vm_working_set_delta_size/smpte_3840x2160_60fps_vp9.webm_renderer

Revision             Result                 N
chromium@452930      247676 +- 799.409      21      good
chromium@453062      250822 +- 44411.8      21      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 media.tough_video_cases

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8982636260417867472

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5249377691500544


| 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!
►

Sign in to add a comment