| Method of moments (MOM) is a mature and commonly used method used to solve the problem of electromagnetic field scattering. But this method needs large quantity of unknowns to simulate electrically-large-size or complicated-structure object. It requires a lot of computation time and higher hardware configuration, a normal PC can hardly satisfy the requirement. To solve these problems, we can use parallel computing techniques to accelerate the solving. In recent years, with the continuous development of the Graphics Processing Unit (GPU), its powerful parallel computing ability attracts a lot of developers to study the GPU-based general-purpose computing. NVIDIA’s CUDA (Compute Unified Device Architecture) programming model enables developers to use the GPU for efficient parallel computing in a familiar development environment and its efficiency is much better than the CPU in the implementation of a large number of computing.This paper studies how to improve the efficiency of MOM for electromagnetic field scattering based on GPU. In the process of MOM, filling the impedance matrix and solving the matrix equation cost a lot of time. The operation of filling the impedance matrix and matrix-vector multiplication in iteration algorithm for solving the matrix equation are well suited for parallel processing. Compared with the traditional computer cluster, the new parallel method based on GPU only needs one computer which equipped with CUDA-enabled GPU, and its efficiency is higher.Firstly, the parallel MOM for solving two-dimensional electromagnetic scattering problems is implemented on the GPU. Then the parallel MOM for solving three-dimensional electromagnetic scattering problems is implemented and optimized with smaller particle size. Finally, this paper achieved the parallelization of the high-order MOM based on the best uniform approximation. The high-order MOM has higher accuracy fewer subdivision points than the low-order MOM, and reduces the computational complexity and memory. |
PDF Full Text Request