[rescue] wild far-out theory.

[Dave McGuire]

On February 7, Bill Bradford wrote:
> Dave M. - any idea what the performance on a Cray-1 was like, in
> MFLOPS/GFLOPS?

  133 MFLOPS.

> I'm having a Joshua Boyd moment here.
> 
> cray-1 emulator.

  Hmm, nice idea.

> (yes, I'm sick and perverted)
> 
> Anybody know if the instruction set, etc, is publicly documented?

  Yup, all that info is available.  I'd say its practical.  The problem
would be getting an OS.  Unicos is a unix variant, but they didn't
always run Unicos...back in the -1 and XMP days, they ran as attached
processors with a front-end system...usually a Data General machine, a
VAX, or an IBM mainframe.  Front-end interfaces for several different
types of systems were available.  You'd write your code in, say, VAX
FORTRAN, link it against the Cray libraries, run the executable on the
VAX, and the heavy math stuff would happen on the Cray, accessed by
the Cray libraries on the VAX via the front-end interface.

    -Dave

-- 
Dave McGuire
St. Petersburg, FL         "Less talk.  More synthohol." --Lt. Worf