On Tue, 10 Oct 2000, ip500 wrote:
This sould just make the 10 year rule.... Ran across
one over the
weekend at a local University auction. Anyone have pointers to Doc's on
these cool Hypercubes?
Damn, I've been looking for one of those :-)
Whatever docs I have are archived on tape and burried someplace. I played
around with an iPSC/1 around 1987, so you're OK as far as the 10-year rule
goes. AFAIK, it was Intel's first attempt at a supercomputer. It was
based on the Cosmic Cube project at Cal Tech.
It's basically a bunch of 286's in a hypercube topology. Each node had
something like 512K RAM, and the internode communication wasn't very fast
-- it might have been ethernet.
I don't recall there being any console as such. You talked to the thing
through a 286 box running Xenix connected via ethernet, which uploaded
your parallel app to each of the nodes, started the app, and collected the
output.
It ran an OS called CrOS, which was basically just a communication
manager. There was no shell or filesystem or anything like that -- the
box was intended to be used as an execution engine. Intel provided a
Unix-hosted simulator (which I think was developed at Cal Tech), so you
could debug the apps using more or less standard tools before uploading
your code.
Cheers,
Doug