Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!usc!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!att!emory!hubcap!stein
From: stein@dhw68k.cts.com (Rick 'Transputer' Stein)
Newsgroups: comp.parallel
Subject: Re: CM-2 SIMD question
Message-ID: <12531@hubcap.clemson.edu>
Date: 9 Jan 91 00:29:44 GMT
References: <12515@hubcap.clemson.edu>
Sender: fpst@hubcap.clemson.edu
Organization: Wolfskill & Dowling residence; Anaheim, CA (USA)
Lines: 25
Approved: parallel@hubcap.clemson.edu

In article <12515@hubcap.clemson.edu> john%ghostwheel.unm.edu@ariel.unm.edu (John Prentice) writes:
>inherently a SIMD type problem.  However, different nodes in our mesh
>use different equations of state.  Some are tabular, some are
>analytic, some are iterative.  Clearly you can't perform all these
>calculations in lock step.  So, I assume we have to turn off all processors
>except those using a specific equation of state, do the EOS calculations for 
>those nodes, and then turn those processors off and turn on those for the next
>EOS, etc... until we have exhausted all the types of EOS we need.  Is there
>a better way to do this?
>John K. Prentice
If I understand SIMD class computation the way I think (and this always
suspect, at least at this time), you have two options:

1) either all cpus execute the same instruction,
2) or, they ignore it.

Hence, you've got to do one set of equations (or, equivalently operate)
on a type of data stream on x processors, feed them a no-op, and then do
the other data stream on y processors, while no-oping the other x cpus.

Apologies if this is b.s. (What can you say, I'm a MIMD-type at heart).
-- 
Richard M. Stein (aka, Rick 'Transputer' Stein)
Sole proprietor of Rick's Software Toxic Waste Dump and Kitty Litter Co.
"You build 'em, we bury 'em." uucp: ...{spsd, zardoz, felix}!dhw68k!stein