[PlanetCCRMA] Ardour not fast enough?

Mark Knecht Mark Knecht <markknecht@gmail.com>
Fri Jan 21 08:11:02 2005


On Fri, 21 Jan 2005 17:02:08 +0800, Timo Sivula <timo.sivula@luukku.com> wrote:
> Hi,
> 
> Ardour and Jack seem to have some quarrels in my machine. With regular
> ingervals, 5-10 minutes, Jack says:
> 
> "subgraph starting at ardour timed out (subgraph_wait_fd=11, status = 0,
> state = Running)"
> 
> Ardour complains that it was fast enough for Jack, and stops reproducing
> sound. To continue I have to restart Ardour.
> 
> I have the same problem with ardour-0.9beta19, 22 and 23. The only
> effect I have in use is a master reverb with Freeverb.
> 
> $ rpm -q jack-audio-connection-kit
> jack-audio-connection-kit-0.99.0-2.rhfc1.ccrma
> 
> I am using the fedora core 1 version of planet CCRMA. Anybody else with
> this problem?
> 
> Timo

Hi Timo,
   I guess one obvious question would be how fast is the machine, but
I doubt that this is caused by a slow machine really.

   I had some problems sort of like this quite awhile back. In my case
they were, we thought, caused by one or more of the following:

1) The APCI system
2) CPU clock frequency scaling
3) Interrupt service latencies (not IRP priority)

At that time I disabled ACPI and frequency scaling and things got much
better. WIth the newer kernels and some of Ingo's patches we got to
the point where you could set specific interrupt service routines to
be threaded or non-threaded. Setting the sound card to be non-threaded
then allowed it to be serviced quickly and I no longer got any timeout
problems in Jack. This was, at the time, a kernel I built myself.

I am now back to running a standard Planet FC2 kernel.
(2.6.10-2.1.ll.rhfc2.ccrma) I haven't noticed any major problems with
this kernel but I haven't pushed it very hard. I'm running it due to
other problems I was having with one of the earlier Planet kernals.
This kernel does not, unfortunately IMO, offer the patch that allows
me to set the threaded/non-threaded irq, but it seems to work fine at
normal latencies, or so I think for now... ;-)

HTH,
Mark