Parallel computing and its impact on computational electromagnetics

It is noted that massively parallel, distributed memory computers will provide the increases in computing performance needed to solve the largest problems in computational electromagnetics. The authors describe such a system, covering its architecture, performance, and programming in relation to a model problem drawn from low-frequency magnetics. The results from this model problem are used to estimate the computing power required in different areas of computational electromagnetics. Finally, the issue of ease of programming, including software issues such as parallel libraries, is discussed