Introduction

The widespread use of comupters in problem solving, in both science and technology, has stimulated the development of many numerical algorithms. Most of these algorithms are based on the reduction of an initial differential probelm, to the same problems in linear algebra. At present this method is the most widely used means for solving applied problems. It is natural that the dimensionality of the linear algebra problems obtained depends on the discretization parameter, the latter being generally the finite difference mesh-size. Thus, more accurate solutions require smaller discretization parameters. By decreasing the parameter we generally increase the number of linear equations, which usually results in an increase in the amount of computation time.