[linux-audio-dev] ANN: Distributed LADSPA Processing

[John Check]

On Thursday 19 August 2004 11:05 am, Nelson Posse Lago wrote:
> On Wed, Aug 18 2004 at 03:22:23pm -0400, Paul Davis wrote:
> > thats potentially a bit ugly, but in practice, its probably going to
> > work, especially with JACK-driven clients/hosts.
>
> That's the idea; in fact, this was mostly triggered by the "but it's
> native processing!" text on ardour's site.
>
> > does the metaplugin have a control output port called "latency"?
>
> Not yet, but that's planned; we just have to add up all the "real"
> plugin's latencies and the latency inserted by the distributed system.
>
> > >If jack udp transport doesn't work adequately [...]
> >
> > jack.rtp is probably a better candidate :)
>
> I believe the uses are different.
>
> DLADSPA was intended to work in a studio to increase the processing
> capability of the system and assumes a LAN that won't fail (which is a
> very reasonable assumption when your timing requirements are above 1ms).
> This means we may keep latencies very low, down to 3ms total in-to-out
> latency if the hardware can handle 1ms periods. Something similar would be
> useful for jack, opening the possibility of running two instances of
> ardour on different machines, for instance.
>

BTW, I met a guy on the train today who does audio for video work.
I'm going to tour his place early next week and he seems like a good candidate
to help test some things. Also, the SSL rated guy gave me a killer application 
for dladspa even if latency sucked. ;)

> OTOH, adding rtp to the mix makes a lot of sense if you intend to cross
> routers or even WANs, but is much more complicated. Low latency is

Long distance transmission is ISDN territory. Interfaces are cheap enough.
A standalone dual channel codec lists for around $5k. There is some merit to 
the idea of a ISDN capable gateway to interface a control room with the rest 
of a network. It really should be on it's own subnet anyway. Bad enough 
people rip CD's, they don't need to break into session archives.

> probably ruled out, but of course networked audio is very interesting on
> its own. But maybe this is trying to reinvent the wheel, given that MAS
> already exists...
>
> See ya,
> Nelson

It only needs to fill a niche. UNIX paradigm, eh?