🚀 go-pugleaf

Peter Brewer of the SRC makes a good point.
I've not tried to say whether loosely coupled distributed networks
constitute supercomputers.

My favorite example is Manasse's factoring using email of all things.
I think its just a matter of time before a few other supercomputer
users will resort to things like using gangs of workstations to solve
their problem.  I know one of our computational fluid dynamicists
has Mark's code and has given serious thought to taking spare cycles
on our workstations rather than use supers.  I think he intends to ask
rather than use an improved Morris style flooding algorithm.

It's also a matter of time for other disciplines to do this as well.
Sure it will be harder than writing convention code, but some users are
getting desparate for their fractions of a super. 10% or 1% of a super
sounds like a lot of what people are getting everywhere.  Imagine
battling worms seeking out unused cycles (one could write a novel, maybe
make a movie)....

But I think for purposes of discussion, supers are confind to "one box"
be it a 2 foot high octogon or a aircraft carrier sized build
(Cray machine to Smith machine).  A different newsgroup can be used for
loosely coupled distributed computing.

As to 3090s being supers.  David just sent me a note: he visited
IBM Japan's Tokyo Research Lab.  Since 3090's aren't seen as supers, I should
just post this to s.c.j. (which most don't read, but I fear the noise
ratio might drive some of you away): use the '=' command and just look
at headers (if rn, or nn, etc.), I might just forward these things to
news groups as needed rather than post to s.c.j., but then I have to
think of which group to post......

--e. nobuo miya, NASA Ames Research Center, eugene@orville.nas.nasa.gov
  {uunet,mailrus,other gateways}!ames!eugene