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

Issue 631061 link

Starred by 2 users

Issue metadata

Status: Duplicate
Merged: issue 630669
Owner:
Last visit > 30 days ago
Closed: Aug 2016
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression

Blocked on:
issue 632712



Sign in to add a comment

7.4%-52.7% regression in system_health.memory_desktop at 406978:407004

Project Member Reported by petrcermak@chromium.org, Jul 25 2016

Issue description

See the link to graphs below.
 
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Jul 25 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@406977  4419271  843819   12  good
chromium@407004  4840000  1167735  8   bad

Bisect job ran on: mac_10_11_perf_bisect
Bug ID: 631061

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests system_health.memory_desktop
Test Metric: load_games-memory:chrome:all_processes:reported_by_chrome:gpumemorybuffer:effective_size_avg/load_games_bubbles
Relative Change: 20.87%
Score: 0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/mac_10_11_perf_bisect/builds/762
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9006152342748543072


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=6650906766999552

| 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!
Trying another bisect.

===== 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@406977  8427520  0.0      12  good
chromium@407004  8400213  77234.9  8   bad

Bisect job ran on: mac_10_11_perf_bisect
Bug ID: 631061

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests system_health.memory_desktop
Test Metric: load_tools-memory:chrome:all_processes:reported_by_chrome:gpumemorybuffer:effective_size_avg/load_tools_docs
Relative Change: 0.52%
Score: 0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/mac_10_11_perf_bisect/builds/813
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9005161195443554608


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5812522096001024

| 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!

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


===== TESTED REVISIONS =====
Revision         Mean      Std Dev  N  Good?
chromium@406977  8499371   792243   5  good
chromium@406999  8877056   1245346  5  good
chromium@407010  8499371   1043026  5  good
chromium@407016  8077995   331301   5  good
chromium@407017  9123669   1813648  5  good
chromium@407018  12160683  97695.3  5  bad
chromium@407019  12291755  249075   5  bad
chromium@407021  12466517  390781   5  bad

Bisect job ran on: mac_10_11_perf_bisect
Bug ID: 631061

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests system_health.memory_desktop
Test Metric: load_news-memory:chrome:all_processes:reported_by_chrome:gpumemorybuffer:effective_size_avg/load_news_nytimes
Relative Change: 46.68%
Score: 0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/mac_10_11_perf_bisect/builds/818
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9004899616195697776


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5869874270699520

| 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!
Blockedon: 632712
r407018 is a skia roll, blocking on the bisect bug about skia rolls.

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


===== TESTED REVISIONS =====
Revision         Mean      Std Dev  N  Good?
chromium@406977  8499371   792243   5  good
chromium@406999  8877056   1245346  5  good
chromium@407010  8499371   1043026  5  good
chromium@407016  8077995   331301   5  good
chromium@407017  9123669   1813648  5  good
chromium@407018  12160683  97695.3  5  bad
chromium@407019  12291755  249075   5  bad
chromium@407021  12466517  390781   5  bad

Bisect job ran on: mac_10_11_perf_bisect
Bug ID: 631061

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests system_health.memory_desktop
Test Metric: load_news-memory:chrome:all_processes:reported_by_chrome:gpumemorybuffer:effective_size_avg/load_news_nytimes
Relative Change: 46.68%
Score: 0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/mac_10_11_perf_bisect/builds/818
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9004899616195697776


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5869874270699520

| 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: bunge...@chromium.org mtklein@chromium.org bsalomon@chromium.org dvonbeck@google.com robertphillips@chromium.org msarett@chromium.org brianosman@chromium.org caryclark@chromium.org herb@chromium.org egdaniel@chromium.org
The bisect blamed a Skia roll (https://codereview.chromium.org/2167973005) for this regression (+2.9 MiB in gpumemorybuffer on NY times). Unfortunately, the bisect didn't descend into the roll (issue 632712).

mtklein,bungeman,brianosman,herb,msarett,bsalomon,caryclark,dvonbeck,robertphillips,egdaniel: Could you please take a look and see if any of your patches might have caused this?
Cc: -mtklein@chromium.org
It's not any of my 4 CLs.
Cc: -egdaniel@chromium.org
Highly doubt its my vulkan change since chrome doesn't use it
Cc: -caryclark@chromium.org
My change is 1 character that has no memory impact
Cc: -msarett@chromium.org
Not me
My 2 CLs from yesterday are only comments and 5 lines of GLSL code, and
it's not code that users can run so that couldn't have been me.
Seeing as how the bisect was done on a Mac... the one change I have in that roll is to code which isn't compiled into the Mac build.
Project Member

Comment 20 by 42576172...@developer.gserviceaccount.com, Aug 10 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@406977  7736932  524552   12  good
chromium@407004  8192523  409982   8   bad

Bisect job ran on: mac_10_11_perf_bisect
Bug ID: 631061

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests system_health.memory_desktop
Test Metric: load_tools-memory:chrome:all_processes:reported_by_chrome:gpumemorybuffer:effective_size_avg/load_tools_maps
Relative Change: 0.57%
Score: 0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/mac_10_11_perf_bisect/builds/820
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9004899616195697776


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5869874270699520

| 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 22 by 42576172...@developer.gserviceaccount.com, Aug 15 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@406977  4382862  807405   12  good
chromium@407004  4077227  113060   8   bad

Bisect job ran on: mac_10_11_perf_bisect
Bug ID: 631061

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests system_health.memory_desktop
Test Metric: load_games-memory:chrome:all_processes:reported_by_chrome:gpumemorybuffer:effective_size_avg/load_games_bubbles
Relative Change: 0.00%
Score: 0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/mac_10_11_perf_bisect/builds/830
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9004240867051266816


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5888707819732992

| 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!
I tried to manually bisect this within the Skia roll. I ran:

tools/perf/run_benchmark -v --browser=release --also-run-disabled-tests system_health.memory_desktop --story-filter=.*load.*new.*nytimes.* 

Across ~15 runs I got zero for the stat system_health.memory_desktop:load_news-memory:chrome:all_processes:reported_by_chrome:gpumemorybuffer:effective_size_avg all but one time.
petrcermak: Any clues why the bisects in #6, #20, and #22 could not reproduce any change? Could the GPUs on bots be configured differently?

bsalomon: This is reproing on one specific perfbot config (a Mac Mini running 10.11), and for some reason it doesn't repro 100% of the time. We are not seeing it on the other perfbots.
sullivan: No idea :-( I'm running another two bisects, one of them on nytimes, where a previous bisect (#8) almost succeeded (but wasn't able to descend into the roll).
The failed bisects ran on the following slaves:

build165-b4, build166-b4

The only successful bisect ran on the following slave:

build165-b4

Given this, it doesn't look like an infrastructure problem.
I haven't found any difference between the bots' GPU configurations in failed and successful bisects.
Project Member

Comment 30 by 42576172...@developer.gserviceaccount.com, Aug 17 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@406977  8182670  196682   12  good
chromium@407004  8173568  182304   8   bad

Bisect job ran on: mac_10_11_perf_bisect
Bug ID: 631061

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests system_health.memory_desktop
Test Metric: load_media-memory:chrome:all_processes:reported_by_chrome:gpumemorybuffer:effective_size_avg/load_media_dailymotion
Relative Change: 0.00%
Score: 0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/mac_10_11_perf_bisect/builds/839
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9004076208396268192


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=6370873867501568

| 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 31 by 42576172...@developer.gserviceaccount.com, Aug 18 2016

Mergedinto: 630669
Status: Duplicate (was: Assigned)

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


===== SUSPECTED CL(s) =====
Subject : Reenable support for CHROMIUM_set_uniform_location
Author  : bsalomon
Commit description:
  
BUG=skia:5015
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2167303002

Review-Url: https://codereview.chromium.org/2167303002
Commit  : d7bae18c7beda8c3ae6fa95c9c8630f30403b9db
Date    : Thu Jul 21 19:49:30 2016


===== TESTED REVISIONS =====
Revision                         Mean      Std Dev  N  Good?
chromium@406977                  8165376   284828   5  good
chromium@406999                  8077995   365542   5  good
chromium@407010                  8121685   195391   5  good
chromium@407016                  8789675   1107706  5  good
chromium@407017                  8077995   249075   5  good
chromium@407017,skia@2895eeb11a  7990613   267550   5  good
chromium@407017,skia@7fbfbbe8f4  8630443   954803   5  good
chromium@407017,skia@a339bb0d95  8411989   706482   5  good
chromium@407017,skia@d7bae18c7b  12248064  182771   5  bad    <--
chromium@407018                  12204373  218453   5  bad
chromium@407019                  12073301  365542   5  bad
chromium@407021                  12248064  447696   5  bad

Bisect job ran on: mac_10_11_perf_bisect
Bug ID: 631061

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests system_health.memory_desktop
Test Metric: load_news-memory:chrome:all_processes:reported_by_chrome:gpumemorybuffer:effective_size_avg/load_news_nytimes
Relative Change: 50.00%
Score: 99.8

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/mac_10_11_perf_bisect/builds/840
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9004076171658842528


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5778033022074880

| 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!
Labels: SystemHealth-Sheriff
Labels: -Performance-Sheriff

Sign in to add a comment