[linux-audio-dev] Plugin APIs (again)

[David Olofson]

On Monday 09 December 2002 15.29, Steve Harris wrote:
> On Mon, Dec 09, 2002 at 02:31:24PM +0100, David Olofson wrote:
> [audio rate controls]
>
> > Well, ok. You can have those *too*. (I'm Santa! :-)
>
> Well, you can obviously "have" them, but I dont think you want to
> have sophisticated support for both. That would make the API
> bloaty, and only nuts like me actually use audio rate control in
> 2002.

Where's the bloat? We need audio ports anyway, right?


> Converting between continuous control and event control is not
> reliable, and kinda removes the point of cont. control.

Yes, but without converters, you can't do things like applying audio 
effects on controls...


> > > I look forward to a few years where all DSP code can run
> > > blockless and with audio rate (or near audio rate) control.
> >
> > Have you considered that timestamped events may actually have
> > *better* than sample accurate timing...? ;-) (Well, I actually
> > suspect that subsample accurate timing will show up sooner or
> > later - as a buzword or because it actually matters. Maybe we
> > should allocate some bits for it while we're at it?)
>
> I dont think its meaningful. At least not efficiently. I't be a
> better use of cycles (and devloper time) to just run the whole
> thing at 96k IMNSHO.

Yes, but that's only one bit per sample. Is that sufficient for the 
most extreme accuracy nuts?

Anyway, no I don't think there would be much point in messing this, 
at least not until we've actually seen it in use, and understand 
better *why* it's used.


[...]
> > I don't think that DSPs or generic CPUs will *ever* be fast
> > enough that you can completely stop worrying about performance.
> > You think convolution is heavy stuff? Wait until you see what the
> > cutting edge
>
> Hey, I've never though that you can forget about performance, its
> just that the cost of things like blockless processing and
> contonuous control get lost in the noise when you start throwing RT
> convolvers around.

Well, yeah - and if you can run hundreds of *those*, it probably 
doesn't matter that every single synth voice spends more CPU time 
processing control data than audio. :-)


> And aynway statments like "CPUs will *ever* be fast enough"... are
> usually proven false later ;)

Yes - but I'd rather not wait ten years before I can actually *use* 
my software! :-)

In fact, I've already waited *more* than ten years already for PCs to 
become at all usable for serious audio synthesis and recording. Now 
they are, but since I didn't have Linux/lowlatency some years ago, I 
never got around to write any hopelessly inefficient software that 
would have been just fine today. ;-)


> There are some hardware synths in existence today that use cont.
> control and blockless processing. The improvement in sound quality
> is noticable.

Do they use that for *everything* (like all parameters, switches 
etc), or just where it actually matters?


//David Olofson - Programmer, Composer, Open Source Advocate

.- The Return of Audiality! --------------------------------.
| Free/Open Source Audio Engine for use in Games or Studio. |
| RT and off-line synth. Scripting. Sample accurate timing. |
`---------------------------> http://olofson.net/audiality -'
.- M A I A -------------------------------------------------.
|    The Multimedia Application Integration Architecture    |
`----------------------------> http://www.linuxdj.com/maia -'
   --- http://olofson.net --- http://www.reologica.se ---