Basic HTML version of Foils prepared May 7 1996

Foil 22 Single Instruction Multiple Data (SIMD)

From MetaComputing -- MRA Meeting Part I:Concepts and Issues Tutorial for CRPC MRA Meeting at Cornell -- May 7 1996. by Mark Baker, Geoffrey Fox


The SIMD architecture consists
Many (typically simple) processors, with some local memory.
Executing the same instruction in lockstep, on a small piece of data in its local memory, with the instructions issued by the controller processor.
Such architectures are good for applying algorithms which require the same operation on a large array of elements, however, they suffer badly if the problem results in load imbalances as the processors synchronise after every step. This is not usually a problem for data parallel programs.
Examples - AMT DAP - Maspar



Northeast Parallel Architectures Center, Syracuse University, npac@npac.syr.edu

If you have any comments about this server, send e-mail to webmaster@npac.syr.edu.

Page produced by wwwfoil on Sun Apr 11 1999