MAPPING FINITE-ELEMENT GRIDS ONTO PARALLEL MULTICOMPUTERS USING A SELF-ORGANIZING MAP

LSOM (load-balancing self-organising map), a neural network based on K ohonen's self-organising map, is proposed for the problem of mapping f inite element method (FEM) grids to distributed memory parallel comput ers with mesh interconnection networks. The rough global ordering prod uced by LSOM is combined with the local refinement Kernighan-Lin algor ithm (LSOM-KL) to obtain the solution. LSOM-KL achieved a load imbalan ce of less than 0.1% and a low number of hops, comparable to results o btained with commonly used recursive bisection methods.