Enhanced Parallel FDTD Method Using SSE Instruction Sets

摘要

To accelerate the simulation of the parallel FDTD method, this paper proposes an effective hardware acceleration technique based on the SSE instruction sets, and puts forward a three-level data parallel algorithm based on MPI, OpenMP and SSE instructions. To demonstrate the acceleration effect of this technique, this paper develops two types of codes using C language: one is based on MPI + OpenMP, another is based on MPI + OpenMP + SSE, and then draws a comparison between the computing time of the two types of codes in the numerical experiments for the same electromagnetic radiation problems. The experimental results show that this acceleration technique can achieve an acceleration rate of 2.44 for the ideal case on a PC cluster and 2.37 for the practical problem on a 2-CPU workstation without requiring any extra hardware investment, and provide an efficient and economical technique for the electromagnetic simulations.