[comp.arch] Questions about Myrias SPS-2

mdr@reed.UUCP (Mike Rutenberg) (11/14/88)

I would love to see a complete description of the architecture of the
myrias machine and some effort to compare it to existing parallel
machines like the CM-2 & Crays.  All the material that I have seen
since 1982, when I first learned of this effort, has been exceedingly
vague.  The main claim has been the existence of "pardo", which acts
like a "do" but specifies that the statements of that block should be
exectuted in a separate process, possibly on a seperate processor.

There seems to be global memory access available to a process with a
hierarchical structuring of fiberoptic communications channels between
processors.  Have any measurements been done on memory contention?

What appplications, of the ones mentioned in the press release, are
currently running and in customers hands?  What is the machine
*particularly* good at?

What does it use as an operating system?

And many more questions...

Mike Rutenberg    Reed College, Portland Oregon   (503)239-4434


---------
cmt@myrias.UUCP (Chris Thomson) quotes the press release:

>"The inspiration for the Myrias system came from physical scientists,
>who have created a programming environment that requires no special
>programming skills," he said.  "Users have been able to convert codes
>to parallel operation on the system in hours."

[...]
>The "pardo" is the only language construct that the user programmer sees in
>order to invoke parallelism.  Parallel Fortran and C subprograms can be
>combined; pardo can be used within recursive subroutines, and can be nested.

[...]
>The SPS-2 is scalable, and can be built up from 64 to over a thousand
>processors to provide supercomputer performance.  The system will be
>effective on a range of applications and methods such as: geophysics,
>molecular modelling, molecular dynamics, image and signal processing, Ising
>models, aerodynamic and hydrodynamic modelling, particle transport models,
>n-body simulations, ray tracing, multi-dimensional transformations and
>convolutions, text processing and retrieval, VLSI design, and others.