Parallel hybrid iteration methods for block bordered linear systems (Q1126588)

Parallel iterative methods are discussed to solve large sparse block bordered systems of linear equations. The iterative methods are based on Gauss-Seidel, successive overrelaxation or accelerated overrelaxation iteration. However, for the pivotal diagonal block, the multisplitting technique is employed to enhance parallelism (more precisely processor utilization). Convergence is proved for linear systems the coefficient matrix of which is an \(M\)-matrix, an \(H\)-matrix, or a symmetric positive definite matrix. The parallel methods are compared by means of a 2-processor simulation with the well-known stationary methods by means of the number of iteration steps needed to solve a simplistic system of equations.