[ardour-dev] First successful test run

[Florian Schmidt]

On Sun, 5 Sep 2004 19:44:00 +0100
Ben Bell <bjb-ardour-dev at deus.net> wrote:

> On Sun, Sep 05, 2004 at 07:54:46PM +0200, Florian Schmidt wrote:
> > > Things still causing problems: 2.6 kernels - xruns galore even
> > > with voluntary
> > >  preempt. rosegarden4<->jack<->ardour integration -- midi timing
> > >  all over the place with fluidsynth, jack.udp.
> > have a look. maybe you get bitten by this..
> Interesting. Actually it turns out I've been running
> LD_ASSUME_KERNEL=2.4.1 for quite a while because of rxvt futex hanging
> problems. I'd forgotten, though.
> Testing with L_A_K=2.4.22 seemed superficially to be better but
> running quite a lot of qsynth channels and thrashing around virtual
> desktops caused overruns eventually. Frankly, it could be better with
> 2.4.22 instead of 2.4.1 but I'm fed up with starting everything up,
> and thrashing it to see if it breaks. If the info would be helpful to
> someone I'll dig a bit more but otherwise I think I'll go back to my
> much happier new box for now ;-)

Hi,

some shots into the dark:

1] you do run jack with the -R option right? If not, does running as
root help? Does running as root help anyways [even if you are already
running with -R but as ordinary user]?

2] have a look at /proc/interrupts. do you see something not alright,
like an ERR count, or your soundcard sharing irq's with something else?
Also you probably use APIC, right? See if running in single cpu mode and
with XT-PIC improves the situation [i never used MP systems so i don't
know if this is possible]..

3] when using voluntary preempt, make sure you have your soundcard irq
set to not threaded and also have hardirq and softirq redirection
enabled [so other irq's don't preempt the soundcard irq]..

4] what periodsize are you using?

flo