DSP vs. Microprocessor

Hi,

Andrew: yours was a really good description of the landscape of
"silicon used to perform Digital Signal Processing".... most of us,
when hear "a DSP" inmediately think of the chips sold by vendors like
TI, ADI or Frescale, that can be called a "general purpos digital
signal processor". And there are many other chips to do digital signal
processing, it all depends on the needs of the application.

JaaC

On Sun, 02 Jan 2005 16:54:22 -0800, jaac wrote:
> Andrew: yours was a really good description of the landscape of
> "silicon used to perform Digital Signal Processing".... most of us,
> when hear "a DSP" inmediately think of the chips sold by vendors like
> TI, ADI or Frescale, that can be called a "general purpos digital
> signal processor". And there are many other chips to do digital signal
> processing, it all depends on the needs of the application.

Indeed.  And the landscape gets more complicated and blurry all the time,
as video games start to do lots of real maths to deal with 3D perspective
and physics simulations, and as even very small applications wind up doing
a little signal processing on the side.  Even PIC processors have some
features to help with DSP these days, and modern ARM and MIPS architecture
variants are positively brimming with features to help with DSP, including
single cycle (or so) multiply-accumulate-with-saturation instructions. 
The PowerPC architecture had a (floating point) multiply-accumulate
instruction from the very beginning.

So just picking on a few architectural peculiarities isn't going to help
with the classification task, IMO.  It's all a question of appropriate
optimizations for the application space.  If the applications are
exclusively "signal processing", then there's a bunch of stuff that one
likes to have, and a bunch of stuff that you can (generally) choose
to leave out (like MMUs, for example).

Here's a few extreme points to think about, though:

Is a TI TAS3103 a DSP?

Is the ARM7 used in (for example) the iPod a DSP?

Is a one-chip flash 8051 or PIC a DSP if it has a bunch of built-in A/D
and D/A converters in it?

Is a Cray X-MP a DSP?  What if it's being used to process a radar
beam-forming problem?

How about an Intel Itanium or DEC Alpha? How about an Intel i860?

What if either of the above three are installed on one or several VME or
cPCI cards and running under VXWorks or other real-time kernel, (rather
than in a Unix or VMS workstation,) and running a radar installation or
CAT scanner?  What about an array of PowerPC or SPARC processors, in the
same configuration? What if the processors in the same application were
SHARCs or TMS320C40s?  Transputers?  (All of these have been available
off-the shelf for these sorts of applications, at one time or another.)

Cheers,

-- 
Andrew

This is actually a pretty frequently asked question in this group.  So it should
not come as a surprise that an answer appears in our very own comp.dsp FAQ!

http://www.bdti.com/faq/3.htm#32

The answer in the official FAQ above goes along with most of the posts in this
thread.

"Brad Griffis" <bradgriffis@hotmail.com> wrote in message
news:_fRBd.8759$_X7.1447@newssvr33.news.prodigy.com...
> I'd say the distinguishing characteristics of a DSP are as follows:
>
> 1)  Single-cycle MAC:  Many DSPs today can even do two MACs in a single
> cycle.
> 2)  Repeat single instruction and repeat block of instructions.
> 3)  Addressing Modes:  Capability to do hardware circular addressing and
> other post-increments/decrements in the same cycle as the instruction.
> 4)  Multiple memory accesses.  Often there are multiple buses available.
>
> It really takes all 4 of the above characteristics to be able to implement a
> super-efficient FIR filter (i.e. a few setup cycles and then 1 cycle per
> tap)
>
> Brad.
>
> "dspguru_wannabe" <gogichen@gmail.com> wrote in message
> news:1104602593.082839.83060@c13g2000cwb.googlegroups.com...
> > What is the most fundamental difference between a DSP (processor) and a
> > Microprocessor? I could list out a few, but what core features make a
> > DSP a DSP? What makes a Microprocessor not capable of performing DSP
> > applications unless emulated by software?