New issue
Advanced search Search tips

Issue 729672 link

Starred by 1 user

Issue metadata

Status: Duplicate
Owner: ----
Closed: Jun 2017
Cc:
EstimatedDays: ----
NextAction: ----
OS: Windows
Pri: 2
Type: Bug-Regression



Sign in to add a comment

2%-17.4% regression in media.tough_video_cases_tbmv2 at 476227:476344

Project Member Reported by chcunningham@google.com, Jun 5 2017

Issue description

Likely a duplicate of  Issue 729630 , but the perf dashboard is not updating that issue as I link additional alerts. So filing a separate issue to ensure we get a bisect. 

Duplicates will work out naturally when the bisect blames the common CL

Uptick in 
- size-in-bytes for allocated objects metrics 
- on win7 video tests. 
 
Mergedinto: 729205
Status: Duplicate (was: Untriaged)

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

Suspected Commit
  Author : Reilly Grant
  Commit : 64bff37c24aefed6abfc5066f8ff5c49e71072f7
  Date   : Thu Jun 01 13:31:01 2017
  Subject: Ship WebUSB

Bisect Details
  Configuration: win_perf_bisect
  Benchmark    : media.tough_video_cases_tbmv2
  Metric       : memory:chrome:renderer_processes:reported_by_chrome:v8:heap:old_space:effective_size_avg/video.html?src_tulip2.wav_type_audio
  Change       : 39.63% | 1323008.0 -> 1847296.0

Revision             Result              N
chromium@476238      1323008 +- 0.0      6      good
chromium@476265      1323008 +- 0.0      6      good
chromium@476266      1323008 +- 0.0      6      good
chromium@476267      1847296 +- 0.0      6      bad       <--
chromium@476269      1847296 +- 0.0      6      bad
chromium@476272      1847296 +- 0.0      6      bad
chromium@476278      1847296 +- 0.0      6      bad
chromium@476291      1847296 +- 0.0      6      bad
chromium@476344      1847296 +- 0.0      6      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 --story-filter=video.html.src.tulip2.wav.type.audio media.tough_video_cases_tbmv2

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8977601502237780848

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=4867657980968960


| 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!
Labels: OS-Windows
I've run this test locally and cannot reproduce this result by either reverting the patch above or flipping WebUSB from stable to experimental.

ToT (WebUSB stable): https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_0-2017-06-07_18-18-09-87798.html

ToT with WebUSB experimental: https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_0-2017-06-08_10-24-00-53536.html

ToT with patch reverted: https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_0-2017-06-08_11-07-18-76634.html

The result in all cases is that memory:chrome:renderer_processes:reported_by_chrome:v8:heap:old_space:effective_size is 2,560.0 KiB. Note, this is different from the results above and I cannot explain that.
Where in the traces should I be looking for effective size? (Sorry, haven't used these much).

I don't know much about WebUSB - not sure how it may be having some effect. These tests are definitely not calling that API. 

I do see that a number of different regressions seem to be blaming your CL, which adds some confidence that there may be an issue. Though its weird that a total revert didn't have any effect on the value. 

I'll start another bisect with a starting point that is ~20 revisions before your CL - if its before you we should see it. 

https://chromeperf.appspot.com/buildbucket_job_status/8977230854554565536

Sign in to add a comment