Applying High Performance Computing in Electromagnetics

Abstract

Parallel processing is now a realistic possibility. There are numerous of-the-shelf High Performance Computing (HPC) platforms available, on which to implement computationally intensive algorithms. HPC can be applied in the field of Computational Electromagnetism to investigate problems that were so computationally expensive that they were practically “unsolvabe”. However, there are two emerging questions in order to solve these really complex electromagnetic problems, firstly, which algorithm to use and secondly, on which HPC platform the algorithm will be executed. The rapid portation and evolution of algorithms, originally devised for sequential machines, to current HPC platforms, are now beginning to reach their limit of performance as predicted and defined by Amdahl’s law.The available HPC platform performances, are now vastly exceeding the expectations imposed by these algorithms. In order to exploit and utilise all the available performances of current and predicted HPC platforms, newly parallel-based algorithms (such as Monte-Carlo based techniques) have to be devised. As presented in this chapter, one such algorithm is the Parallel Method of Moments (PM0M) technique. The resulting algorithm parallelisation enables for the proposed approach to be applied in the domain of Electromagnetics, to analyse electrically large planar conducting structures on various diverse computing platforms. In order to determine which HPC platform we will use for future applications with even larger problem sizes, a comparison is made of the platforms applicability/suitability, ease of code porting and performance obtained.