Cooperating sparse simplex algorithm for a distributed memory multiprocessor

Abstract

We undertake a computational analysis of the algorithmic components of the sparse simplex (SSX) method and consider the performance of alternative SSX solution strategies. We then combine these alternative strategies within an asychronous control structure of a parallel algorithm, implemented on a distributed memory multiprocessor machine. Our experimental results not only show (limited) speedup and robust performance, they also give new insight into the internal operation of the SSX and its adaptation on a parallel platform.