X-Message-Number: 26907
Date: Wed, 31 Aug 2005 10:42:45 -0400
From: Thomas Donaldson <>
Subject: parallelism, for Daniel Crevier

Hi everyone!

A bit for Daniel Crevier:

The point of my argument is that we CANNOT imitate N processors working in
parallel with one that works N times faster. The problem doesn't show
up at all if we're just doing scientific computations, where there's
no problem at all in doing that. The problem shows up (theoretically
at least) if we made a robot able to imitate a living thing, and 
shows up again when each processor in the N is doing something 
different from the others which must be done AT THE SAME TIME. Think
about the movements you would be making to flee a tiger, for instance.
You cannot wait until your right leg has completed its motion, or 
even part of its motion, before you move both your arms and your
left leg. 

In theory, again, if the processor is sufficiently fast, we could
break up all the motions involved in fleeing a tiger so that it
would go through each one and then do the next. However that just
"solves" one particular problem, and the solution breaks down if
the many motions involved must be even faster for a fixed processor,
or again if so many motions are needed that the single fixed processor
cannot deal with ALL of them in a short enough time. Even if we 
suppose that we can make our single processors arbitrarily fast,
we can also provide problems for them too big for them to solve
in the needed time. (There is likely to be a limit as to how fast
we make a processor, but I'm discussing an even more basic problem
which lies behind this).

All this is why people became interested in parallel computers
in the first place. I'll add that I got interested in parallel
computers in computer science as a mathematician, and one fact
about parallel computing is that it can require different algorithms
than for a sequential computer, plus programming features meaning-
less in sequential computers. And as an immortalist, I don't
believe that even the ability to wait forever for an answer
will make us decide not to use parallel computers. And yes, as
I said to Yvan, I think that TIME should take a much more 
prominent role in computer thinking than it does now.

           Best wishes and long long life,

                 Thomas Donaldson

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=26907