81567 - image size (perhaps transparency) impacts javascript performance

Status: RESOLVED WORKSFORME

(Core :: Layout, defect)

Description

[Chris Casciano]

noticing greatly improved performace when executing javascript in the most
recent nightly [Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:0.9+)
Gecko/20010517] I went back through some old javascript experiments I have on my
site. In doing so I noticed that the following 2 pages, that vary *only* in size
of the image contained in the moving layers, actually move at tremendously
different speeds.

http://placenamehere.com/libraryDoodles/20010104.02moz.html runs at a speed
roughly equal to what i see in IE5.5/PC as well as the speed i recollect it
running at last time i looked at the page on a Mac. This page has images that
are 3px squares with no transparency

http://placenamehere.com/libraryDoodles/20010104.03moz.html runs much, much more
slowly. The html code and javascript on this page is *identical* to the first
page in every way except that the image contained in each div is this:

<img src="oval.gif" width=400 height=200 alt="" border=0>

a 400x200 gif, 1.92k, with some transparency.

--

Now, I know there are plenty of performace bugs files already that are dupes of
dupes of dupes... I just knew of no other logical place to convey this finding

Comment 1

[Phil Schwartau]

Browser, not engine. Reassigning to Layout for further triage.

Confirming bug on WinNT 2001051710 nightly. The second testcase
runs much more slowly than the first, and much more slowly than
it does in NN4.7.

The behavior on Linux 2001051708 seems different. In this case,
the first testcase does not seem to load properly at all
(I can't see a single image, and slider bars keep appearing and
disappearing [both horizontal and vertical]). The second testcase
seems to be about as slow on Linux as it is on WinNT - 

Comment 2

[Chris Casciano]

just to satisfy my own curiosity I will make and post a 3rd version of this code
using a 400x200 NON-transparent version of that ovalimage to see if the
tansparency of the image has anything to do with it. 

[that is if no one else does before i do]

Comment 3

[doctor__j]

I've checked with System Monitor and both testcase suck up the same amount of
CPU juices (which is pretty high in fact - at around 75% while the normal CPU
kernel usage level is around 30%)

Comment 4

[Chris Casciano]

ok, sorry for the bit of a delay... i have just created the following:

http://placenamehere.com/libraryDoodles/20010104.04moz.html

with this image contained in each div is this:

<img src="ovalfilled.gif" width=400 height=200 alt="" border=0>

a 400x200 gif, with no transparency.

It seems that the use of transparency is the major contributing factor in the
animation slowing down. On my comp there is a slight variation in speed between
the small solid images moving and this new version. Its the version with the
large transparent image that is lagging *way* behind the other two.

Comment 5

[Chris Casciano]

I'm curious to know if this performance issue is getting looked at at all. It's
been over a month and I see no signs of activity in bugzilla, and no noticeable
improvement as a side effect of other work. Something I find kinda strange when
I compare it to my other experiences with this project.

Comment 6

[anthonyd]

the performance of the moving images has definitely gotten better (with todays 
build) but still, when compared to IE, we get our butts kicked.

This is not a table specific bug anyhow, reassigning to core owner.

adding perf keyword.

Comment 7

[Chris Casciano]

how would one go about getting this listed under the dhtml performance tracking bug?

http://bugzilla.mozilla.org/show_bug.cgi?id=21762

Comment 8

[Kevin McCluskey (gone)]

Reassigning to Don.

Comment 9

[Chris Waterson]

Is this still a problem? I'm not noticing much of a difference between
02moz.html and 03moz.html on Linux.

Comment 10

[Chris Casciano]

Hmmm.... yeah, this is noticably better on 200203309/Win2k... and now that i'm
looking at it again 0.9.9 as well... Was there a related checkin somewhere?

Comment 11

[Chris Casciano]

just going through some of my old bugs... looking at the test pages in 1.0/Win2k
I'm happy with the current performance. Not sure if this is attributed to any
one patch, or just an accumulation of a bunch of performance fixes. I'm willing
to mark resolved/worksforme. any objections?

Comment 12

[Corrado Berti]

working good (Comparable to IE6);
Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.1a) Gecko/20020619

Comment 13

[Brian Haskin (Janzert)]

Works for me also, marking as such.

Comment 14

[Markus Hübner]

confirming.
what would be interesting is to use a gif that is resized (scaled) by Mozilla.
I think this will considerably slow down things.
this can be seen in bug 117436.