152171 - Mozilla doesn't scale to multiple processors

Status: RESOLVED INCOMPLETE

(Core :: XPCOM, defect)

Description

[Rene Pronk]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.1a) Gecko/20020614
BuildID:    2002061408

Although Mozilla uses multiple threads, it doesn't scale at all to multiple
processors. I have a dual smp box running Mozilla and no matter what I do, I
can't get the cpu usage above 50%, which means that it doesn't actually use more
than 1 processor. When I load e.g. a flash animation that uses all the cpu time
it can get (http://www.matrox.com/mga/) in a tab and load it again in another
tab, I would suspect both my cpu's to run at full speed. It stays at 50% however
(== 1 cpu). Although this is just plain bad, it gets even worse when I open yet
another tab and try to load a normal page in it. It seems that Mozilla uses the
same thread again or does some sort of blocking that prohibits using multiple
threads efficiently, because it takes a lot of time to render the page, even
though I have a full processor left waiting for something to do.

Note that this is not only a performance problem on smp systems, but also on
single cpu systems, since plugins are apparently able to block Mozilla from
doing what is has to. Even the gui suffers from this because the system get less
responsive.

Reproducible: Always
Steps to Reproduce:
1. Open http://www.matrox.com/mga/ in first tab
2. Open http://www.matrox.com/mga/ in second tab
3. Open another site in third tab

Actual Results:  Everything works, but terribly slow. GUI gets less responsive

Expected Results:  Everything works, but should be a lot faster since I have a
cpu left waiting for a job. gui should respond normal.

Comment 1

[Markus Hübner]

confirming using trunk build 2002062508 on win-xp.

Comment 2

[Hixie (not reading bugmail)]

Moving all threading bugs to XPCOM. See bug 160356.

Comment 3

[filidautore]

Is there a chance to improve this situation?
This seems a good snappy bug... or I missing something?

Comment 4

[Josh Matthews [:jdm]]

I don't think this bug really has any value in its current state. It's too general for Snappy, but doesn't actually have any meta dependencies. We have other bugs covering multi-process use, and specific bugs about ways we can improve threading.