289681 - Slow performance in real-world newsticker

Status: VERIFIED DUPLICATE of bug 137584

(Firefox :: General, defect)

Description

[Ilya Konstantinov]

I've received many complaints from friends about slow performance on
www.ynet.co.il (a major Israeli news site) front page. I've managed to
narrow-down those complaints to their news ticker:
http://www.ynet.co.il/Ext/Comp/Ticker/Dhtml_Ticker/0,10016,L-184,00.html

This newsticker, implemented entirely in JS, uses two positioned objects to
create an endless ticker. Every 40ms it moves a 116px text up by 1 pixel. This
brings our CPU usage *way* too high.

(IE, apparently, stomaches this ticker just fine.)

Comment 1

[Ilya Konstantinov]

Attachment: Testcase altering the position of a text box by 1px every 40ms

Comment 2

[Steve England [:stevee]]

Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.8b2) Gecko/20050408
Firefox/1.0+
For me (AthlonXP2400) and the testcase, CPU usage stays around 15% - firefox
remains responsive at all times.
Ilya, what version of firefox are you using, and do you still see the issue with
a firefox nightly (1.1beta) build?
http://ftp.uni-erlangen.de/pub/mozilla.org/firefox/nightly/latest-trunk/

Comment 3

[Ilya Konstantinov]

On my Athlon 1000, it's around 40%, which does makes the rest of the system a
bit more slugish. Furthermore, having the viewport receiving those updates
completely obscured doesn't lower the CPU usage. Yes, Nightly Firefox builds
don't help much with this CPU usage either.

Our gauge for this bug should be about having performance comparable to IE,
which is what people complained about in the first place.

Comment 4

[José Jeria]

With a nightly build (20050408) the url in commment 0 uses 20% CPU usage. The
attached testcase uses 5% of the cpu usage.

Comment 5

[Martijn Wargers (dead)]

See bug 260340 and bug 137584 and bug 235694.
This one should also be marked as a duplicate of bug 137584, not?

Comment 6

[José Jeria]

I agree with comment 5, duping.

*** This bug has been marked as a duplicate of 137584 ***

Comment 7

[Steve England [:stevee]]

Should the testcase be attached to bug 137584, seeing as it doesn't have one?

Comment 8

[José Jeria]

(In reply to comment #7)
> Should the testcase be attached to bug 137584, seeing as it doesn't have one?

The url testcase would be better (20% cpu usage), than the testcase (below 5%).

Comment 9

[Ilya Konstantinov]

The testcase can be spruced up with some tables and other slow-rendering
features. It can also be made wider (e.g. 400px) -- that'll also slow it down.
However, 5% vs. 40% is not the point. It should hardly take any CPU time, just
as it does on IE.