SOLUTION OF LARGE LINEAR SPARSE SYSTEMS B Y PARALLEL ITERATIVE METHODS

The Conjugate Gradient method is widely used for solving large sparse linear systems. Its parallelization shoes the prohibitive cost of the reduction operation (which consists of computing the optimal parameter s at each iterations). In this paper, we propose an analysis of the al gorithm derived from the Conjugate Gradient method, with constant para meters. We remind the link with the second order Richardson method and compare it with the Conjugate Gradient method. Then, we study their p arallelizations, and propose an implementation on the Connection Machi ne.