[linux-audio-user] "Linux fx box how to" WAS: some thoughts about Linux audio software documentation

[Russell Hanaghan]

On Fri, 2004-08-13 at 16:15, Lee Revell wrote:
> On Fri, 2004-08-13 at 18:57, Russell Hanaghan wrote:
> > > >
> > > > My question is this...how useful will it be in the big scheme of things?
> > > > I think Ive only read posts from like 2 or 3 other people that even
> > > > wanted to do this. Most seem to be creating music or recording it. I do
> > > > that to but not very much lately and thats not what this system is set
> > > > up for.
> > > 
> > > Don't let other peoples apparent disinterest stop you. A good idea is a good 
> > > idea. Don't think about the people who are up on things, think about the 
> > > people who could use a killer rack on stage and have a used PC kicking 
> > > around.
> > 
> > I won't...thanks for the encouragement. I started with just that; a
> > spare pc and experimenting. Now granted it is only as my PA fx right now
> > but that is due only to lack of time and I can only experiment so much
> > before I approach the boundaries of danger...IE; Not having the box up
> > for the next gig!
> > 
> 
> My primary interest in Linux audio involves using it in a live setting. 
> This is a lot closer to a hard realtime situation than recording, where
> you at least have a chance of trying it again.  I think it's a much more
> interesting problem.  It's also one that other general purpose OS'es
> have absolutely no chance of solving.
> 
> The 2.6 kernel with Ingo's voluntary preemption patches should allow
> hard realtime type guarantees for audio use.  As long as the ALSA driver
> is realtime safe (should be the case), and all interrupts except the
> soundcard are threaded (doable), and all jackd clients are also realtime
> safe (already required), and assuming no bugs (still working on it),
> then we have a hard realtime system.  The types of timing guarantees we
> can provide would have to be determined by testing.
> 
> Lee
> 
The upside to using sound processing only as opposed to recording to HD
is one substantial device is eliminated. It seems that many folks
headaches with Xruns, etc, stem from HD access. Also graphics...

As far as 2.6 goes...I'm getting reasonable performance from 2.6.7 with
lsm. Mainly I use the Realcap GID=29 to get RT as user. It is still
kinda Xrunny but mostly due to graphics based stuff it seems. I'm using
Xfce4 a lot which has a lite footprint and good functionality. 

It would still be nice to get the kernel thing sorted. I actually read
on the mdk cooker list a post that Svetlo (i think) said the Voluntary
patches are included and enabled by default in his kernels...How might I
determine that?

R~