PARALLEL WAVEFORM-NEWTON ALGORITHMS FOR CIRCUIT SIMULATION.

A number of relaxation-based simulation algorithms have been developed to reduce runtimes on uniprocessor computers, including an algorithm based on the waveform-Newton technique which has been implemented in the SPLAX program. To obtain further runtime improvement, a number of parallel algorithms are developed for the waveform-relaxation-Newton method. In particular, variations of the Gauss-Jacobi and Gauss-Seidel methods are explored. Also included is a novel approach that attempts to perform the solutions at different time points in parallel. Results of the implementation of these algorithms in the PSPLAX program are reported.