Time-Domain Computational Electromagnetics Algorithms for GPU Based Computers

Time-domain computational electromagnetic algorithms such as FDTD and TLM require computers with superb processing power and large memory capacity. Grid computing network, cluster computer and massively parallel supercomputers have been the hardware of choices for running powerful modelling tools based on these methods. As a result, high performance modelling tools are only available to elite groups of researchers and big corporations. Stream computing, a new technology that harnesses the tremendous numerical processing power of advanced graphics processing units for general purpose numerical computation, is going to bring high performance time-domain modelling tools to the EM community. This paper reviews the emerging GPU technologies and programming models. Two modelling examples are also used to illustrate the suitability of GPU computing for time-domain electromagnetics.