Parallel processing improvements for full-wave electromagnetic solvers

As computing resources become increasingly parallel for both shared and distributed memory systems, computational electromagnetic methods need to take full advantage of new architectures in order to reduce simulation times. Several different aspects of the solution process make it difficult for traditional finite element field solvers to effectively utilize distributed and shared memory resources. This paper will provide insight into recent advancements made in the finite element method matrix solve (shared memory) and frequency sweep (distributed memory) that drastically reduce simulation times for signal and power integrity applications.