Path: utzoo!censor!geac!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!udel!sbcs!ameristar!rick
From: rick@ameristar (Rick Spanbauer)
Newsgroups: comp.sys.amiga
Subject: Re: Amiga DSP Boards.
Message-ID: <1990Nov27.020113.13352@ameristar>
Date: 27 Nov 90 02:01:13 GMT
References: <1990Nov20.161034.25281@sisd.kodak.com> <1990Nov21.013237.4623@ameristar> <1390@mpirbn.mpifr-bonn.mpg.de>
Organization: Ameristar Technology, Inc
Lines: 28

In article <1390@mpirbn.mpifr-bonn.mpg.de> p554mve@mpirbn.UUCP (Michael van Elst) writes:
>In article <1990Nov21.013237.4623@ameristar> rick@ameristar (Rick Spanbauer) writes:
>>	* Which DSP?  The major players (AT&T, Motorola, TI, Analog Devices)
>>		      all have differing word lengths, instruction sets,
>>		      architectures, chip resources, etc.
>>	o Standard I/O resources?  Should the board have 14, 16, or 18 bit
>>		      a/d, d/a?  What sampling rate (variable)?  What
>>		      pre/post filtering?  A telephone interface for modems?
>>	o IPC model?  Shared memory?  Mailboxes?  Serial interface?
>>	o Memory available?
>
>Well, if these parameters were fixed you won't benefit from future
>enhancements of todays DSPs. I think the major point would be defining
>data structures (f.e. file formats) that you can agree on storing
>and processing information.
>Another point would be an application interface for standard functions
>but you'd need board specific routines for low-level operations.

	DSP's are yet fast enough that they can run some standard threaded 
	language or interpreted instruction set that would abstract the 
	basic machine in a useful way.  Much the same way as the Amiga is 
	wedded to the 680x0 instruction set, a standard DSP must be settled 
	on if we are ever to see a rich application set emerge.

>Michael van Elst

					Rick Spanbauer
					Ameristar