[fluid-dev] Voice renderer, and fluid_voice_t

[David Henningsson]

So my thoughts about the voice renderer and the state machine continues
- what I think now is the correct way to fix Sven's and everybody else's
reordering issues.

I would like to split fluid_voice_t into two parts. One part should be
hard real-time capable and should contain fluid_voice_write and
fluid_voice_mix, and everything that it needs to know to do just that.
But nothing more. That way we try to make the upper bound of CPU time as
low as possible.

In particular, I would like everything related to generators and
modulators to be on the soft real-time side, and only the outcome of
those calculations to move into the hard real-time side.

To get to know the voice code better, I've done some refactoring which
is already in trunk. And as the number of source files increase, so does
the need for subdirectories. Any objections to me organizing the source
files in subdirectories?

// David