A comparative study of two wavefront implementations of a LU solver algorithm

Abstract

Cellular architectures may be used to achive speedup of certain algorithms. However, when there are several ways of mapping the same algorithm onto a cellular array producing approximately the same speedup other measures of effectiveness such as efficiency should be considered in order to compare the quality of different approaches. This paper presents two ways of mapping a LU-decomposition algorithm, both with the same speedup factor but with different efficiencies.