Stream processing approach for computational electromagnetics on graphics processing units

Summary form only given. One often needs to perform repetitive numerical simulations of systems with only one of the parameters being modified, e.g. electromagnetic design optimization or finite temperature modeling of magnetic memories for calculating an ensemble of magnetization state realizations. Computing the realizations is typically done by multiple runs on a single or multiple computing devices. However, such simulations have an overhead of preprocessing, various matrix operations, and excessive data storage.