Subject: Re: [linux-audio-dev] audio application mixing/routing arch
From: Paul Barton-Davis (pbd_AT_Op.Net)
Date: Thu Mar 30 2000 - 06:44:25 EEST
>Has anyone here (I'm new to the list) discussed the harnessing of
>dedicated DSP hardware? (I would like to see the ability to use
>Capybara, Creamware, or the m56002 on my TB Pinnacle for dedicated
>synthesis and/or effects.) This can make all this much easier.
It might make the computational burden a little easier to deal with,
but it introduces a whole series of other problems that I would hardly
think of as "easier". It would certainly be nice to use a dedicated
DSP for certain things, but given that this requires
cross-compilation, inter-processor synchronization, coprocessor run
time environment and a host of other issues, its non-trivial, to say
the least.
About 10 years ago, I wrote the device drivers and the application for
a system that used a very fast DSP chip (for the time) as a dedicated
co-processor under Interactive Unix (a 386 System V Unix system). We
used to run PostScript and a custom halftoning algorithm on it. The
thing worked (actually, it was great!), but I would hardly want to go
back to that kind of arrangement if I could avoid it. The Athlon, the
Alpha and the Altivec all make it look as if this may be
unnecessary. OK, the Altivec is cheating, since it *is* a coprocessor,
but its handled at a level much closer to the host CPU than the stuff
in the Capybara, a Pulsar/SCOPE or your TB card. More like the FPU
associated with an x86.
--p
This archive was generated by hypermail 2b28 : Thu Mar 30 2000 - 13:28:04 EEST