On the convergence of the SAOR method and the error bounds for its acceleration (Q1112552)

The symmetric accelerated overrelaxation (SAOR) method is an iterative method similar to the symmetric successive overrelaxation (SSOR) method, but it differs from that better known method in that it requires the use to choose two rather than one acceleration parameter. If the two parameters are equal, the SAOR method is the same as the SSOR method. In this paper, the authors develop a theory similar to the SSOR theory for the SAOR scheme. Among other matters, they consider the use of semi- iterative methods to accelerate the convergence.