Find in Library

In this study, a fast and accurate method to predict the radar cross-section (RCS) of large-scale and complicated shape targets is proposed based on a high-performance parallel finite difference time-domain (FDTD) numerical method. To this end, several most popular parallel computation methods [including OpenMP, graphics processing unit (GPU), and message-passing interface (MPI)] are discussed first. Based on this discussion, a novel MPI–OpenMP–GPU hybrid parallel computation scheme for FDTD is developed. Moreover, the corresponding load-balance parallel configuration is discussed as well. Since this hybrid parallel scheme combines the merits of existing parallel technologies, the computation performance is remarkably improved. The results show that the computation time of the RCS simulation of a large-scale target can be reduced from 3 days to 0.8 h, that is, ∼98.9% time saving.