Using unified parallel C to parallelize the FDTD method of computational electrodynamics

This paper discusses a parallel implementation of the finite-difference time-domain (FDTD) method using unified parallel C (UPC). UPC is an extension of ANSI C-primarily designed for programming distributed memory parallel hardware. The FDTD method is a powerful and widely used numerical technique in computational electromagnetics community. It is a resource - CPU cycles and memory-hungry technique. Consequently, a parallel implementation is highly desirable to decrease time-to-solution and be able to study larger problems with limited resources. To the best of our knowledge, this is the first attempt to exploit UPC to parallelize the FDTD method. We evaluate the performance of our implementation on a cluster of four SunFire v890 compute nodes - each equipped with 16 processing cores - and report impressive speedups.