高阶辛时域有限差分算法(SFDTD)与传统的时域有限差分算法(FDTD)相比具有更优的稳定性和计算精度,但在进行电磁仿真时则更为耗时。为解决这一问题,文章应用SFDTD的空间并行性,研究并实现了基于计算统一设备架构(CUDA)的SFDTD的并行算法仿真;基于费米架构,分析了各种尺度网格下速度的提升,与传统的CPU实现该算法进行比较,验证了该方法的正确性和高速性。 Compared with the traditional finite difference time domain (FDTD) algorithm, the high-order symplectic finite difference time-domain (SFDTD) algorithm has better stability and calculation accuracy, but it is more time-consuming in electromagnetic simulation. In order to solve this problem, this paper uses the spatial parallelism of SFDTD to research and implement the parallel algorithm simulation of SFDTD based on CUDA. Based on the Fermi architecture, this paper analyzes the speedup of SFDTD under various scale grids, The traditional CPU realizes this algorithm to compare and verify the correctness and high speed of this method.