[linux-audio-user] Re: [Jackit-devel] 2.6.10

[Lee Revell]

On Fri, 2004-12-31 at 00:07 +0100, Florian Schmidt wrote:
> On Thu, 30 Dec 2004 22:40:12 -0000 (WET)
> "Rui Nuno Capela" <rncbc at rncbc.org> wrote:
> 
> > I have to agree with Fernando when he says that when a RT configured
> > kernel boots (and those IRQ threading gets properly tuned) its The
> > linux-audio-dream come true. It just makes everything else look like a
> > joke ;)
> > 
> 
> Full Ack! I'm still rubbing my eyes in disbelief trying to understand
> why there's still people involved with linux audio who haven't tried the
> RP kernels yet. I was dreaming of this sort of functionality since i
> first installed a slackware distribution back in the days (with some 1.x
> kernel) ;)

Yes, of course the RT kernel is a lot better, but I agree with Jack that
it's important to have good realtime support in the base kernel.
Previous Linux kernels were much worse than the proprietary
alternatives, so having a kernel release that's as good or better is a
huge milestone.  Can you imagine Linux owning the server space the way
it does now, if you had to patch the kernel to get superior performance?
Linux got huge in this area because you could take it and install apache
out of the box on a spare machine and get better server performance than
the proprietary guys.

Anyway, everyone's goal is to get the performance of the RT kernel out
of the stock kernel anyway.  So we should figure out what else needs to
go upstream before 2.6.11.  I think in order for it to get much better
we need the softirq and hardirq threading stuff upstream and turning the
might_sleeps into preemption points - IOW the original aspects of VP.

I would still like to see a comparison between RT and DESKTOP.  I think
it's reasonable that we could get PREEMPT_DESKTOP like performance out
of the vanilla kernel very soon.

Lee