Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!mips!apple!voder!nsc!amdahl!JUTS!duts!kls30
From: kls30@duts.ccc.amdahl.com (Kent L Shephard)
Newsgroups: comp.sys.next
Subject: Re: Can DSP's multitask?
Message-ID: <41qC021l083S01@JUTS.ccc.amdahl.com>
Date: 7 Jun 91 14:58:42 GMT
References: <1991Jun4.211914.623@math.ucla.edu>
Sender: netnews@ccc.amdahl.com
Reply-To: kls30@DUTS.ccc.amdahl.com (Kent L. Shephard)
Distribution: na
Organization: Amdahl Corporation, Sunnyvale CA
Lines: 78

In article <1991Jun4.211914.623@math.ucla.edu> barry@pico.math.ucla.edu (Barry Merriman) writes:
>Lately I've gotten into having lots of sonic stuff on
>my NeXT, and noticed that (no surprise) the DSP can only 
>do one thing at a time.
>
>With the advent of products that make heavy use of the DSP
>(like MIX Inc's forthcoming digital modem/fax/answering machine
>all done in the DSP), it seems that the DSP will often be
>occupied when some other process needs it. Is it possible for 
>DSP's to do some sort of multitasking (I guess this would be
>up to the operating system).

The DSP does not run any part of the OS.  It can communicate with the host
processor through a port.  The beauty of the DSP is that you can write
your own code and download it to the DSP.

>
>What I can see in the future, for a single user, is this:
>simultaneously, one could be
>
>(1) listening to music through the DSP

Assuming 44.1 Khz this leaves roughly 11usec to do stuff between samples.

>(2) generating system beeps, voice alerts, etc through the DSP

>(3) number crunching on the DSP (FFT, raytracing, linear algebra)

This is a CPU hog.

>(4) have a modem emulator running on the DSP
>(5) be sending/recieving a fax on the DSP
>(6) have a digital answering machine using the DSP

What you want to do is run 5, count them 5 realtime applications on the
+ 1 more that does not have to be real time.
Well the Quintprocessor will do this.
I doubt that the NeXT will ever do this without several DSPs.

>
>So it seems like the ability to handle around 6 concurrent
>processes on the DSP would be desirable (of course, a slightly
>faster DSP may be needed to really do this).

This is not a feasable feat for 1 DSP.  You are going to need 3 serial
serial ports just to support the telephone stuff.   You are going to need
more than just the small host interface to the '040.

Finally you just are going to need more than 1 DSP to do this.

Also where are you going to store all this DSP code??  The DSP can only
access a very limited memory.  This is not a NeXT architecture problem
but a limitation on the DSP itself.

>
>Hopefully NeXT will have the first OS that includes multitasking
>for coprocessors; Is this possible for the DSP?

Not if you plan to write you own real time code.
You want NeXT to write a multitasking kernal for the DSP?
I don't.    The support they have in the OS is pretty good.  Besides the
DSP is not a full u-processor.  I does what it does very well and that is
not running a full fledged OS.

>
>--
>Barry Merriman
>UCLA Dept. of Math
>UCLA Inst. for Fusion and Plasma Research
>barry@math.ucla.edu (Internet)   barry@arnold.math.ucla.edu (NeXTMail)



--
/*  -The opinions expressed are my own, not my employers.    */
/*      For I can only express my own opinions.              */
/*                                                           */
/*   Kent L. Shephard  : email - kls30@DUTS.ccc.amdahl.com   */