This does sound a little bit like a moral imperative to me!  Nowadays 'the
computer' is already a multi-processor object, even without audio - both the
graphics and disk subsystems employ powerful processors, dsps and local memory,
so that the operating system can concentrate on its primary task - these days,
switching applications. The 'control interface' is message-based, and the
transmission of these messages is managed entirely by the OS, and there is no
guarantee that any message will arrive at its destination at a precise instant
of time. Even the PC timers are not that accurate! I do not believe that any
general-purpose pre-emptive multi-tasking computer can reliably deliver the
essential requirement for latency-free real-time synthesis and performance -
that everything be run within a tightly clocked single-sample loop, which is how
all dedicated hardware-based synthesizers and effects units operate. Even
Extended Csound starts to struggle when ksmps = 1.

The difference, of course, betweeen audio and the graphics + disk subsystems is
that the functionality of the latter two is very tightly defined, with standard
interfaces (such as OpenGL, SCSI, etc), whereas the requirements for audio are
in a fairly constant state of flux and evolution.  This is the justification for
the software approach (indeed Csound  might conceivably amount to the audio
counterpart to OpenGL) - but for any given approach a dedicated hardware
implementation, without the overhead of a multi-tasking OS, will be more
efficient, more reliable, and quite possibly more practical in live performance.
The advantage of the software approach is flexibility, but arguably not speed.

I consider that for us, the current paradigm of the combination of software
synthesis coupled to an optimized hardware-based audio subsystem, which can
itself be enhanced or upgraded as requirements change, is close to optimal, for
the general-purpose computer, and is likely to remain so for a while yet. The
paradigm shift represented by systems such as KYMA ,ISPW and now aparrently
brought more into the mainstream by  the SCOPE system and, potentially, the ADI
card,  is that all the audio software which we presently write and use on a
general-purpose computer can now be run alternatively on a hardware subsystem
designed expressly for the job, without losing any of our cherished flexibility
- it is simply one computer controlling another.

As for the moral imperative, I suspect that this is overruled by a greater one,
that however powerful and fast an audio system, a musician will, sooner or
later, demand more of it that it can deliver!

Richard Dobson


Michael Gogins wrote:

> This is a very intriguing development about which I have definite comments.
>
> In the first place, I firmly believe that the future of music lies strictly
> in software, the only hardware being the computer itself, the control
> interface, the digital to analog converter, the amplifier, and the
> loudspeakers.
>
> This is because it takes much less time to write software than it does to
> create chips and firmware, and regular PCs are now fast enough to do
> considerable DSP and synthesis in real time.
>