New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.
Starred by 2 users

Issue metadata

Status: WontFix
Owner:
Last visit > 30 days ago
Closed: Oct 2012
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Linux
Pri: ----
Type: Bug-Regression



Sign in to add a comment
link

Issue 32223: Perf regression in linux-release-hardy startup_test around r37272

Reported by ericu@chromium.org, Jan 14 2010 Project Member

Issue description

http://build.chromium.org/buildbot/waterfall/builders/Linux%20Perf%20(1)/buil
ds/4979

At revision 36189 we got a perf revision in the linux_startup test.
The most likely culprit is the webkit roll at 36184, but all of 36184-36189 
were new in that build.

startup_test [startup_test
PERF_REGRESS: warm/t t: 162 (132)
] [148 seconds]

Stdio attached.
 
perf.htm
14.3 KB View Download

Comment 1 by atwilson@chromium.org, Jan 14 2010

Looks like this might have been a flaky result since it's green now.

In my webkit roll, there was only one suspicious CL: http://trac.webkit.org/changeset/53196 so it might be 
worth taking a look at that if this perf regression recurs.

Comment 2 by ericu@google.com, Jan 14 2010

It happened again at build 4991 [revision 36228]: PERF_REGRESS: warm/t t: 156 (125).

Comment 3 by tony@chromium.org, Jan 28 2010

There's a lot of variance on the linux perf builder-- I see times between 136ms and 
180ms.  Maybe the we should increase the range before the bot goes red?

I logged into the machine to see if anything strange was running on it that might cause 
the variance, but nothing stuck out.  I went ahead and stopped apache, but that 
shouldn't matter.

Comment 4 by chasechr...@gmail.com, Jan 28 2010

Summary: Perf regression in linux-release-hardy startup_test around r37272
@tc: Thanks for trying to reduce the variance in the tests and for cc'ing me.  It 
appears to me the regression alarm is firing because the startup test actually 
regressed recently.

Re: the initial bug, last week I reverted r36138.  This brought the linux-release-hardy 
startup test back to its previous expectations.  I suspect r36138 was the real cause of 
the initial bug the report covers and r36189 was a revision where the results happened 
to vary enough to trigger the threshold indicator for a perf regression.  Here's the 
link to those results if you want to learn more: 
http://build.chromium.org/buildbot/perf/linux-release-hardy/startup/report.html?
history=50&rev=36200

There's a new linux-release-hardy startup regression.  From what I could read of the 
graph (hard to read), r37272 (a webkit roll), r37357-r37360, r37388 (another webkit 
roll) stood out to me.  I'll update the bug summary to reflect the current state of 
this issue.

Comment 5 by chasechr...@gmail.com, Feb 17 2010

Labels: Performance
Happened again yesterday: PERF_REGRESS: warm/t t: 175 (135)

Comment 6 by chasechr...@gmail.com, Feb 24 2010

Status: Assigned
@ericu, Do you have any cycles for this bug this week?  If not, let's mark this WontFix 
and increase the expected delta in the expectations.

@tc, Agreed about increasing the range.  Let's do that if we WontFix this.

Comment 7 by chasechr...@gmail.com, Feb 26 2010

Labels: -Area-Undefined Area-Internals Regression
Bumping the startup expectations for Linux today to quell the regression bell.

Comment 8 by bugdro...@gmail.com, Feb 26 2010

The following revision refers to this bug:
    http://src.chromium.org/viewvc/chrome?view=rev&revision=40160 

------------------------------------------------------------------------
r40160 | chase@chromium.org | 2010-02-26 13:21:00 -0800 (Fri, 26 Feb 2010) | 7 lines
Changed paths:
   M http://src.chromium.org/viewvc/chrome/trunk/src/tools/perf_expectations/perf_expectations.json?r1=40160&r2=40159

Adjust startup and vista intl expectations.

BUG= 32223 ,  36930 
TEST=startup, vista intl tests are green
TBR=nsylvain@chromium.org

Review URL: http://codereview.chromium.org/661202
------------------------------------------------------------------------

Comment 9 by c...@chromium.org, Apr 14 2010

Comment 10 by karen@chromium.org, Jun 24 2010

Labels: karen624move Mstone-X

Comment 11 by lafo...@chromium.org, Mar 18 2011

Labels: -Regression bulkmove Type-Regression
http://build.chromium.org/buildbot/waterfall/builders/Linux%20Perf%20(1)/buil
ds/4979

At revision 36189 we got a perf revision in the linux_startup test.
The most likely culprit is the webkit roll at 36184, but all of 36184-36189 
were new in that build.

startup_test [startup_test
PERF_REGRESS: warm/t t: 162 (132)
] [148 seconds]

Stdio attached.

Comment 12 by lafo...@chromium.org, Mar 18 2011

Labels: -Performance Stability-Performance
http://build.chromium.org/buildbot/waterfall/builders/Linux%20Perf%20(1)/buil
ds/4979

At revision 36189 we got a perf revision in the linux_startup test.
The most likely culprit is the webkit roll at 36184, but all of 36184-36189 
were new in that build.

startup_test [startup_test
PERF_REGRESS: warm/t t: 162 (132)
] [148 seconds]

Stdio attached.

Comment 13 by ericu@chromium.org, Oct 30 2012

Status: WontFix
Way obsolete--Chase loosened the expectations two years ago.

Comment 14 by bugdroid1@chromium.org, Mar 9 2013

Project Member
Labels: -Area-Internals -Type-Regression -Stability-Performance Type-Bug-Regression Cr-Internals Performance

Sign in to add a comment