Newsgroups: comp.parallel.pvm
From: mapdjb@midge.bath.ac.uk (D J Batey)
Subject: Re: MASPAR
Organization: School of Mathematical Sciences, University of Bath, UK
Date: Wed, 6 Jul 1994 10:38:43 GMT
Message-ID: <CsILKK.GzA@midge.bath.ac.uk>

Steve Heistand writes:
> Ok, one more machine question, has anyone used the MASPAR with PVM?
> It compiles nicely and runs, but it seems only to be running on
> the front end.

That's correct -- the message passing paradigm presented by PVM is not
appropriate to SIMD machines such as the MASPAR; the processing
elements only run in lock-step, directed by the front-end, all active
processors executing the same instruction on their local data (hence
SIMD - Single Instruction, Multiple Data). So there is no question of
asynchronous point to point communications with or between the
processing elements of a SIMD array.  The only reasonable use for PVM
in these situations is to connect the front-end to the virtual
machine, so that other components can send messages directing the
front-end to use the processor array.

-----------------------------------------------------------------------
 Duncan Batey                                          Computing Group,
 djb@maths.bath.ac.uk                            School of Mathematics,
 Postgraduate Research Student             University of Bath, England.
-----------------------------------------------------------------------




