[PD] a/d/a and jackd alsa low-latency problems

Frank Barknecht fbar at footils.org
Wed Oct 20 10:13:21 CEST 2004


Hallo,
Matju hat gesagt: // Matju wrote:

> On Tue, 19 Oct 2004, Frank Barknecht wrote:
> 
> > Whooho, 64 already is *very* small. I never heard of anybody run jack
> > with 32. That's almost hard-realtime!
> 
> Hi. I'd like to know what you consider to be hard-realtime. I was under
> the impression that it meant that the problem requires strict time limits
> on the execution, and therefore using pd audio in a way that causes no
> dropouts may already be considered as hard-realtime, no matter how big the
> latency.

I didn't intend to be technically correct, but illustrative. The
correct description of hard realtime is of course yours: fixed upper
latency limits. In that way, Linux might be considered hard-realtime
already, with a hard latency of, say, several minutes. ;) Generally
though people are not interested in hard latencies that big.

I was more referring to the special purpose realtime OSes, some based
on Linux like RTLinux, which actually provide hard realtime in the
microsecond area.

For our normal uses this would be overkill, IMO, and it's not the goal
of most efforts in getting lower latencies for the general purpose
Linux kernel. Although there are efforts to get hard realtime into the
kernel, too. [1]  This is important for several kinds of embedded
devices, and the results of course can benefit our needs, too. But
then probably an overhaul of Pd's architecture would be needed to
make use of it. 

[1] http://source.mvista.com/linux_2_6_RT_Announce.txt

Ciao
-- 
 Frank Barknecht                               _ ______footils.org__




More information about the Pd-list mailing list