🚀 go-pugleaf

wells_s@kosmos.wcc.govt.nz (Steve Wells) writes:

>OS/2 does NOT HAVE 'TRUE' multitasking!
>Multi-tasking is running 2 or more processes AT THE EXACT SAME TIME!
>You can't do this without a chip to do it, or without halving the
>speed of the machine for each extra process!  (e.g. every program
>you wish to multi-task!)

This is a use of the term `multi-tasking' with which I was not
previously acquainted.  In fact, the entire field of Computing Science
is not aquainted with it.  You are describing multi-processing.  Have
you considered taking a few CS courses?  Just basic survey courses
would be helpful, by the sound of your posts.  They're easy credits;
you know, `hacks for jocks' (apologies to geology's `rocks for
jocks' courses).

BTW, multiprocessing with heterogenous processors is generally not
much good in general purpose computing environments.  It gets too
difficult to optimally distribute the threads of a job when they're
architecturally dependant.  All the really whizzy multi-processors
are homogenous; for instance, big Crays and the Connection Machines.

Now as regards the claim that the DSP chip is a big performance
enhancement -

The idea that you can get big gains *in the general case* from
specialized hardware is often referred to as "Pike's Inequality", and
has been tried (and subsequently abandoned) repeatedly in the history
of computing.  Spending the transistor budget on the CPU(s) always
seems to win out.  Come on over and lurk in comp.arch to get a feel
for the issues.

On the other hand, I cheerfully agree that special-purpose machines
(simulators, controllers, dsp's, games boxes, & c) can get useful boosts
by optimizing their architecture for a restricted problem domain.

regards,
Ross

--
Ross Alexander  rwa@cs.athabascau.ca  (403) 675 6311  ve6pdq@nebulus.ampr.ab.ca
	"All television is children's television." -- Richard P. Adler