Waveform methods for space and time parallelism (Q1184117): Difference between revisions

Some convergence properties as well as considerations on the parallelism of general iterative methods (generalized waveform relaxations) for the numerical solution of ordinary differential equations are studied. The methods under consideration are generalized Picard iterations and they are mainly addressed to solve large systems of ordinary differential equations that arise in circuit simulation. Firstly, after introducing a measure of the rate of convergence termed as speed of convergence, some convergence results for Jacobi- and Gauss-Seidel-type relaxations are given. Finally, some considerations on the parallelism across space and time of these algorithms are given.