2003-11-17 03:42:10

by Stan Bubrouski

[permalink] [raw]
Subject: Bad interactivity with 2.6-test9

Hey,

I have been testing 2.6-test9 (Arjan van de Ven's 2.6.0-0.test9.1.67
RPMS rebuilt) tainted by nvidia drivers.

The kernel has worked awesome and desktop interactivity works great
under most circumstances, however I have noticed something ratehr
annoying.

When I start watching a video under mplayer and have something like
spamassassin doing a sa-learn in an x-term if I start the sa-learn
before mplayer and then play a video, the sa-learn makes no progress
until I switch away from mplayer (playing full screen). After which
sa-learn continues to filter through messages much faster, and even
while mplayer is going is still progressing much faster than before I
switched away from mplayer the first time.

Very odd. Con anyone ideas?

-sb


2003-11-17 18:54:42

by Rik van Riel

[permalink] [raw]
Subject: Re: Bad interactivity with 2.6-test9

On Sun, 16 Nov 2003, Stan Bubrouski wrote:

> When I start watching a video under mplayer and have something like
> spamassassin doing a sa-learn in an x-term if I start the sa-learn
> before mplayer and then play a video, the sa-learn makes no progress
> until I switch away from mplayer (playing full screen). After which
> sa-learn continues to filter through messages much faster, and even
> while mplayer is going is still progressing much faster than before I
> switched away from mplayer the first time.
>
> Very odd. Con anyone ideas?

Looks like scheduler starvation. Con ?

--
"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it." - Brian W. Kernighan

2003-11-17 20:04:40

by Stan Bubrouski

[permalink] [raw]
Subject: Re: Bad interactivity with 2.6-test9

On Mon, 2003-11-17 at 13:54, Rik van Riel wrote:
> On Sun, 16 Nov 2003, Stan Bubrouski wrote:
>
> > When I start watching a video under mplayer and have something like
> > spamassassin doing a sa-learn in an x-term if I start the sa-learn
> > before mplayer and then play a video, the sa-learn makes no progress
> > until I switch away from mplayer (playing full screen). After which
> > sa-learn continues to filter through messages much faster, and even
> > while mplayer is going is still progressing much faster than before I
> > switched away from mplayer the first time.
> >
> > Very odd. Con anyone ideas?
>
> Looks like scheduler starvation. Con ?

Seems like it. I'll do some testing later to see exactly how
consistently I can reproduce this. I have noticed this numerous times
all with -test9.

-sb

2003-11-17 20:07:11

by Con Kolivas

[permalink] [raw]
Subject: Re: Bad interactivity with 2.6-test9

On Tue, 18 Nov 2003 07:03 am, Stan Bubrouski wrote:
> On Mon, 2003-11-17 at 13:54, Rik van Riel wrote:
> > On Sun, 16 Nov 2003, Stan Bubrouski wrote:
> > > When I start watching a video under mplayer and have something like
> > > spamassassin doing a sa-learn in an x-term if I start the sa-learn
> > > before mplayer and then play a video, the sa-learn makes no progress
> > > until I switch away from mplayer (playing full screen). After which
> > > sa-learn continues to filter through messages much faster, and even
> > > while mplayer is going is still progressing much faster than before I
> > > switched away from mplayer the first time.
> > >
> > > Very odd. Con anyone ideas?
> >
> > Looks like scheduler starvation. Con ?
>
> Seems like it. I'll do some testing later to see exactly how
> consistently I can reproduce this. I have noticed this numerous times
> all with -test9.

Usual checks. See if it happened with test6,7,8 as well as 9.
Do `vmstat 1' output to file and `top -d 1' output to file. They can tell if
it's a dynamic priority related scheduler starvation issue.

Con