[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: Parallel Processing & Symbolics

>> From: DMittman@SARG-URANUS.jpl.nasa.GOV
>>2)	I am tossing around the idea of using a Symbolics
>>	as a front-end to a Hypercube Mark IIIfp (parallel
>>	machine). Is there anything you would like to see
>>	accomplished in this domain?
>>3)	What is the current standard (semi-standard) for a
>>	Parallel LISP, and where can I get a copy?

There is no standard for Parallel Lisp.  In fact, I believe
there is a lot of misunderstanding and confusion about
what a parallel Lisp is.
The Lisp that runs with the Connection Machine is considered
a parallel Lisp, but it is a world apart from the Lisp
that runs on a distributed-memory Intel Hypercube, and once again, 
a world apart from the Lisp that runs on a shared-memory Encore Multimax.
Programming a network of Symbolics can also be seen as ``Parallel Lisp''.
Different parallel architectures have different ways of doing
things, with different costs and parallelism overheads,
with a corresponding difference in the way you express the parallelism.

What it means to be a ``front end'' is also unclear.
Top Level, Inc. has a debugging and development interface that runs on a 
Symbolics for developing parallel programs running on the Sequent Symmetry or
Encore Multimax shared-memory multiprocessors.  But the Symbolics
machine is not strictly neccessary to run code on the multiprocessor.
The Symbolics ``front end'' for the Connection Machine is a much
different thing.  A *LISP program is a program that actually runs on the Symbolics,
but uses the Connection Machine as a ``co-processor''.