[OT] Re: [linux-audio-dev] Linux 2.6 not a latency panacea?

[Takashi Iwai]

At Thu, 14 Aug 2003 16:27:37 +0200,
Robert Jonsson wrote:
> 
> Hi,
> 
> > > I also found 2.6.0-test[123] to be less responsive than 2.4.x-ll, or even
> > > stock 2.4.x. I've also experienced XMMS dropouts under load (for example
> > > compiling Muse)
> > >
> > > Some behaviour I've noticed is that under heavy load the desktop/audio
> > > doesn't freeze for a certain block of time, but rather in short (~2
> > > seconds) intervalls...
> >
> > this should have been imporved significantly in -mm tree.
> > even reducing the min/max timeslices would help a lot.  the default
> > values look too large for desktop users...
> 
> A fairly of topic question: Since 2.6 contains several schedulers; is this 
> selectable at runtime or is it a compiletime switch ?

it's defined statically in kernel/sched.c:

/*
 * These are the 'tuning knobs' of the scheduler:
 *
 * Minimum timeslice is 10 msecs, default timeslice is 100 msecs,
 * maximum timeslice is 200 msecs. Timeslices get refilled after
 * they expire.
 */
#define MIN_TIMESLICE           ( 10 * HZ / 1000)
#define MAX_TIMESLICE           (200 * HZ / 1000)
...

(remember that HZ=1000 in 2.6 for i386.)

> Sounds shaky to select at runtime, but infinitely cool :)

i guess it would be relatively easy to replace the above with
variables controlled via sysctl.


Takashi