🚀 go-pugleaf

In <74042@cup.portal.com> Bob_BobR_Retelle@cup.portal.com writes:

>Steve Wells writes:
>[A lot of good applications for a Digital Signal Processor]

>Except..

>>Co-Processor.  Programs should be able to use it as a secondary processor
>>to speed up a lot of operations.  (e.g.  A MUCH enhanced blitter etc.)

>A DSP has a VERY limited instruction set, highly optimized toward
>multiply and shift operations, the kinds of things you'd want
>to do to a stream of real-time signals (ie: audio or video)

>It CAN'T do the things a math-coprocessor can do, so its use
>as a "co-processor" would be extremely limited.

If one wants a math-coprocessor, it  certainly  makes  much  more
sense  to  add an FPU to the Falcon rather than pulling some hack
to use the DSP for math stuff. I do think, however, that the  DSP
could   be  used  for  many  purposes  for  that  no  specialized
coprocessors exist. As an example, from what  I  know  about  the
DSP,  it would be possible to use it to compute the edit distance
between two strings (that is, how many times you have to  type  a
key  at  least in order to transform one string to the other one,
casually speaking).
In this sense, the Falcon's may point toward a  future  evolution
of   computer   architecture   --   they   may   actually  become
"heterogeneous transputer systems", having an array of  different
processors,  each of which can be loaded with code supplied by an
application to fulfill a specialized task with high performance.

Maybe just a weird idea...

Greetinx, Jan

 +- Jan Kim -- X.400:    S=kim;OU=vax;O=mpiz-koeln;P=mpg;A=dbp;C=de -+
 |             Internet: kim@vax.mpiz-koeln.mpg.dbp.de               |
 |                                                                   |
 *----=<  hierarchical systems are for files, not for humans  >=-----*