Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!ut-emx!nather
From: nather@ut-emx.uucp (Ed Nather)
Newsgroups: comp.arch
Subject: Re: new instructions
Keywords: fixed, fractional
Message-ID: <49278@ut-emx.uucp>
Date: 20 May 91 22:11:35 GMT
References: <9105200213.AA05095@ucbvax.Berkeley.EDU> <1991May20.121708.5023@dartvax.dartmouth.edu>
Organization: The University of Texas at Austin; Austin, Texas
Lines: 22

In article <1991May20.121708.5023@dartvax.dartmouth.edu>, mac@eleazar.dartmouth.edu (Alex Colvin) writes:
> >Fixed point arithmetic is little used now because the hardware to support
> >it reasonably well does not exist.  
> 
> Would anyone who works with DSPs comment on this?  They seem to be built for
> fixed point and fractional arithmetic, with the ALU connected to the registers
> through shifters.

For people who use computers for real-time data acquisition and control,
and there are a lot of them, floating point operations are very seldom
used, if at all.  Floating point is valuable for number-crunching jobs,
but running stepping motors or adjudicating multiple interrupts doesn't
need it, and really can't use it.

People use computers for many different jobs, and it's not surprising that
different jobs need different architectures.  What's sort of amazing is how
well our "general purpose" computer architectures do so many different
things quite well indeed.

-- 
Ed Nather
Astronomy Dept, U of Texas @ Austin