Closed Bug 331385 Opened 18 years ago Closed 18 years ago

DHTML perf regression on the "zoom" test on 2006-03-15

Categories

(Core :: Layout, defect)

Product:
Core
Component:
Layout
Platform:
x86
Linux
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED FIXED

People

(Reporter: bzbarsky, Assigned: dbaron)

References

(
URL
)

Details

The "zoom" DHTML regressed by 30% or so on March 15.  Compare the times in http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey/1142434560.16763.gz and http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey/1142447880.24496.gz -- the former has:

  zoom:          748    748     748,748,747,748,750

while the latter has:

  zoom:          979    975     946,979,986,979,986

Current times for "zoom" are a little better (http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey/1143049080.13329.gz):

  zoom:          934    927     895,940,934,935,933

but still regressed.

The initial regression happened in the following bonsai range:

http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=all&branch=HEAD&branchtype=match&dir=&file=&filetype=match&who=&whotype=match&sortby=Date&hours=2&date=explicit&mindate=2006-03-15+06%3A20&maxdate=2006-03-15+10%3A50&cvsroot=%2Fcvsroot

If you're on the CC list and you're not roc, then you have a checkin in that range...

Note that the range is a few minutes too wide on either side, probably.

My bets would be on bug 192767, bug 96394, or bug 318116.

I've modified the "zoom" test so that it outputs the time when run standalone (see URL field).  Unfortunately, I can't reproduce this slowdown locally.  :(
I was planning to look into this today (I did a build-by-date last night right before these checkins).
Assignee: nobody → dbaron
I can't reproduce locally either, based on builds by date across just those three checkins, with a GTK2 build.
Joining the "can't reproduce" crowd:
Comparing the previous nightly, and a tinderbox build from right after these checkins, and running the test 30 times on each of them, I get the following averages:
Before: 724.7
After : 740.8
Which is about 2.2% percent increase - not even close to 30%, and probably statistically insignificant.

(this is all on a G5 Mac)
I can't reproduce in a GTK1 build either.

Note also that the Mac running Tdhtml didn't show an increase, although it was down for a few days when this happened (and it came back faster, although there could have been accumulated problems on the system slowing it down beforehand).
It looks like the fix for bug 331458 fixed this.
Status: NEW → RESOLVED
Closed: 18 years ago
Depends on: 331458
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.