[Jack-Devel] Compiling jack2 in mixed mode

[Fons Adriaensen]

On Sat, Jul 02, 2016 at 02:04:07PM -0400, Paul Davis wrote:

> 4-8 core systems are more or less standard. So this only reduces the cost
> in the best possible scenario by a factor of 4-8. In more common cases, the
> improvement is much less.
 
But in practice still significant.

> > 2. Linear chains could be in the same process. In that case,
> >    depending on implementation, you don't even need a thread
> >    switch.
> >
> 
> Actually, I misspoke - this is already implemented. Internal clients will
> be executed in the same thread if they are part of a serial chain.

But the same doesn't happen (in current Jack implementations) if you
have a linear chain of clients in the same external process. 
 
> IMO, the reality is that chaining together entirely independent signal
> processing code is best done via plugins (internal clients, in a JACK
> context), not processes. The clever hack that JACK represents is useful for
> connecting small numbers of processes, but I don't believe that the design
> should be used as an indication that process-level chains are the right way
> to composite signal processing elements.

So that leads to a two-layer design: a small number of Jack
clients (each of which can be a separate process), and within
each of those a second layer that runs arbitrary graphs of
modules (which can be plugins), in multiple threads.

The problem with this is that the outer layer clients are
always run as a single blob - at the boundary between
the two layers all dependencies are 'ANDed' together, only
because the outer layer can't see the internal graph and
has to assume that all outputs depend on all inputs.

To avoid this, the two layers need to be aware of each other
in some way. It's exactly this that I've been working on. 
The internal scheduler becomes part of the user-side client
library, and interacts with the outer one. 

The result is for example that an external insert (as in
Ardour) doesn't add latency. Of course, in pathological
cases this can lead to more process switches. In practice
that's quite often not the case - the process which 'calls'
the external processor will be able to continue if it has
independent subgraphs (as would be the case for a mixer
with many channel strips).

Ciao, 

-- 
FA

A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)