FDTD Algorithm for Bistatic RCS Prediction of 3-D Target on two GPUs

Abstract

As a numerical algorithm, the finite-difference time domain (FDTD) is effective in solving electromagnetic scattering problem of medium with high complexity. The computional efficiency is low by the traditional central processing unit (CPU) platform. Therefore, the GPU-based FDTD method used to speed up its computational efficiency for bistatic RCS prediction of 3-D object is realized in this paper. Both the validation and efficiency of our implemen is verified by comparison of parallel result versus CPU's. A speedup of about 38x is realized on two NVIDIA K40 GPUs, which improves the computational efficiency. The results also show that the computional efficiency of the parallel method is related to the number of Yee cells.