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

Issue 702596 link

Starred by 1 user
Status: Duplicate
Owner:
jgruber@chromium.org
Closed: Mar 2017
Cc:
jgruber@chromium.org
rmcilroy@chromium.org
█ fran...@chromium.org
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

1%-1.7% regression in memory.top_10_mobile at 457430:457491

Project Member Reported by rmcilroy@chromium.org, Mar 17 2017 
See the link to graphs below.
Project Member

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


=== BISECT JOB RESULTS ===
Bisect was unable to run to completion

Please try rerunning the bisect.


If failures persist contact the team (see below) and report the error.


Bisect Details
  Configuration: android_one_perf_bisect
  Benchmark    : memory.top_10_mobile
  Metric       : memory:chrome:all_processes:reported_by_chrome:v8:effective_size_avg/foreground/http_www_baidu_com_s_word_google

Revision             Result                  N
chromium@457449      4025708 +- 1645.08      6      good
chromium@457491      4096533 +- 3578.31      6      bad

Please refer to the following doc on diagnosing memory regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/memory-infra/memory_benchmarks.md

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests memory.top_10_mobile

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8984869061037926960

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=4576074542874624


| 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 5 by 42576172...@developer.gserviceaccount.com, Mar 23 2017

Cc: fran...@chromium.org
Owner: fran...@chromium.org

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

Hi franzih@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 : franzih
  Commit : 18c35e4958be6a70acc923bf10363eb9aaee5ce4
  Date   : Thu Mar 16 09:25:23 2017
  Subject: Collect type profile for DevTools

Bisect Details
  Configuration: android_one_perf_bisect
  Benchmark    : memory.top_10_mobile
  Metric       : memory:chrome:all_processes:reported_by_chrome:v8:effective_size_avg/foreground/http_www_baidu_com_s_word_google
  Change       : 1.22% | 4041645.33333 -> 4090968.0

Revision                           Result                  N
chromium@457449                    4041645 +- 63265.9      9      good
chromium@457460                    4027241 +- 3409.44      6      good
chromium@457465                    4033628 +- 41476.2      6      good
chromium@457468                    4038655 +- 42347.9      6      good
chromium@457468,v8@3db32e064f      4060188 +- 28434.7      6      good
chromium@457468,v8@5cc6189677      4066303 +- 3643.74      9      good
chromium@457468,v8@9e827c2384      4070988 +- 43134.3      9      good
chromium@457468,v8@460ba9c224      4088710 +- 29907.9      6      good
chromium@457468,v8@18c35e4958      4104443 +- 28014.3      6      bad       <--
chromium@457468,v8@10500cb79c      4099932 +- 23744.8      9      bad
chromium@457469                    4096050 +- 2855.75      6      bad
chromium@457470                    4091790 +- 27373.9      6      bad
chromium@457491                    4090968 +- 31833.9      6      bad

Please refer to the following doc on diagnosing memory regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/memory-infra/memory_benchmarks.md

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests memory.top_10_mobile

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8984382476691746560

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=4576074542874624


| 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 6 by rmcilroy@chromium.org, Mar 24 2017 

Franziska, could you take a look please. Looks like the type profile CL increases memory use even when type profiling isn't enabled (that wasn't the intention, right?).

Comment 7 by fran...@chromium.org, Mar 24 2017 

We assure with a DCHECK that we only add a slot to the feedback vector, when we run with --type-profile. 

The only thing that I see, is that wee add a FeedbackSlot variable to every Assignment in ast.h. That was taken out again in https://codereview.chromium.org/2764113002/diff/40001/src/ast/ast.h

Could that be it?

Comment 8 by rmcilroy@chromium.org, Mar 24 2017 

Hmm, most of the graphs have gone down, however they reduced on a CL range completely unrelated to the type profile code (there was no change around https://codereview.chromium.org/2764113002). I'll kick off another bisect to see if it was definitely caused by type profile feedback.
Project Member

Comment 12 by 42576172...@developer.gserviceaccount.com, Mar 25 2017

Cc: jgruber@chromium.org
Owner: jgruber@chromium.org

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

Hi jgruber@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 : jgruber
  Commit : 5cc6189677066382dd206a68375b88e80ccd9afd
  Date   : Thu Mar 16 07:54:53 2017
  Subject: [regexp] Port RegExpExecStub to CSA (mostly)

Bisect Details
  Configuration: android_webview_nexus6_aosp_perf_bisect
  Benchmark    : memory.top_10_mobile
  Metric       : memory:webview:all_processes:reported_by_chrome:v8:effective_size_avg/background/after_https_www_google_co_uk_hl_en_q_science
  Change       : 1.28% | 3115984.44444 -> 3155980.44444

Revision                           Result                  N
chromium@457449                    3115984 +- 81964.7      9       good
chromium@457466                    3113979 +- 80651.1      9       good
chromium@457468                    3123135 +- 130670       14      good
chromium@457468,v8@3db32e064f      3126349 +- 153623       21      good
chromium@457468,v8@7b76d52083      3129054 +- 80243.5      9       good
chromium@457468,v8@365492f17e      3129457 +- 80341.3      9       good
chromium@457468,v8@9224d5d1bc      3129607 +- 78916.6      9       good
chromium@457468,v8@5cc6189677      3149447 +- 6457.72      9       bad       <--
chromium@457469                    3158531 +- 111883       14      bad
chromium@457470                    3147801 +- 9773.51      9       bad
chromium@457474                    3166657 +- 105752       9       bad
chromium@457482                    3155980 +- 80556.7      9       bad

Please refer to the following doc on diagnosing memory regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/memory-infra/memory_benchmarks.md

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-webview --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests memory.top_10_mobile

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8984234624513406576

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=4926066126225408


| 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 13 by 42576172...@developer.gserviceaccount.com, Mar 25 2017 


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

Suspected Commit
  Author : jgruber
  Commit : 5cc6189677066382dd206a68375b88e80ccd9afd
  Date   : Thu Mar 16 07:54:53 2017
  Subject: [regexp] Port RegExpExecStub to CSA (mostly)

Bisect Details
  Configuration: android_webview_nexus6_aosp_perf_bisect
  Benchmark    : memory.top_10_mobile
  Metric       : memory:webview:all_processes:reported_by_chrome:v8:effective_size_avg/background/after_https_www_google_co_uk_hl_en_q_science
  Change       : 1.59% | 2006648.0 -> 2038470.66667

Revision                           Result                  N
chromium@457449                    2006648 +- 20379.8      6      good
chromium@457466                    2006667 +- 31161.8      6      good
chromium@457468                    2001101 +- 15190.2      6      good
chromium@457468,v8@3db32e064f      2010891 +- 19214.2      6      good
chromium@457468,v8@7b76d52083      2005472 +- 0.0          6      good
chromium@457468,v8@365492f17e      2005472 +- 0.0          6      good
chromium@457468,v8@9224d5d1bc      2007625 +- 15190.2      6      good
chromium@457468,v8@5cc6189677      2035912 +- 0.0          6      bad       <--
chromium@457469                    2038471 +- 19214.2      6      bad
chromium@457470                    2035697 +- 15190.2      6      bad
chromium@457474                    2038471 +- 19214.2      6      bad
chromium@457482                    2038471 +- 19214.2      6      bad

Please refer to the following doc on diagnosing memory regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/memory-infra/memory_benchmarks.md

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-webview --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=after_https_www_google_co_uk_hl_en_q_science memory.top_10_mobile

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8984234609181520128

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5816604509601792


| 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, Mar 25 2017 


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

Suspected Commit
  Author : jgruber
  Commit : 5cc6189677066382dd206a68375b88e80ccd9afd
  Date   : Thu Mar 16 07:54:53 2017
  Subject: [regexp] Port RegExpExecStub to CSA (mostly)

Bisect Details
  Configuration: android_webview_nexus6_aosp_perf_bisect
  Benchmark    : system_health.memory_mobile
  Metric       : memory:webview:all_processes:reported_by_chrome:v8:effective_size_avg/load_news/load_news_hackernews
  Change       : 1.51% | 2103646.0 -> 2135342.66667

Revision                           Result                  N
chromium@457448                    2103646 +- 76.6812      6      good
chromium@457465                    2100887 +- 15190.2      6      good
chromium@457468                    2103632 +- 96.9948      6      good
chromium@457468,v8@3db32e064f      2104827 +- 19287.1      6      good
chromium@457468,v8@7b76d52083      2104979 +- 30426.5      6      good
chromium@457468,v8@365492f17e      2107753 +- 15236.5      6      good
chromium@457468,v8@9224d5d1bc      2107259 +- 15159.8      6      good
chromium@457468,v8@5cc6189677      2141090 +- 102.879      6      bad       <--
chromium@457469                    2138116 +- 96.9948      6      bad
chromium@457470                    2132583 +- 30365.1      6      bad
chromium@457474                    2138088 +- 96.9948      6      bad
chromium@457482                    2135343 +- 15159.8      6      bad

Please refer to the following doc on diagnosing memory regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/memory-infra/memory_benchmarks.md

To Run This Test
  src/tools/perf/run_benchmark -v --browser=android-webview --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=load.news.hackernews system_health.memory_mobile

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8984234598622392192

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5889653313372160


| 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 jgruber@chromium.org, Mar 27 2017

Mergedinto: 703118
Status: Duplicate (was: Untriaged)
The code size regressions of this CL are offset by stubification in https://chromium.googlesource.com/v8/v8/+/10f7b7ecc686b34d13ac3fa5f789dfce1c963711.
►

Sign in to add a comment