Title :
Fast electromagnetic simulation by parallel MoM implemented on CUDA
Author :
Ming Fang ; Kai-Hong Song ; Zhi-Xiang Huang ; Xian-liang Wu
Author_Institution :
Key Lab. of Intell. Comput. & Signal Process., Anhui Univ., Hefei, China
Abstract :
The keys to electromagnetic simulation of arbitrary configuration of conducting surfaces by method of moments (MoM) are filling of impedance matrix elements and solving of linear equations. A CUDA (Compute Unified Device Architecture) enabled graphics processing unit (GPU) launched by NVIDIA company to accelerate implementation of the fast filling of impedance matrix in MoM based on Rao-Wilton-Glisson (RWG) basis functions was presented. Parallel LU decomposition method is applied for the solving of linear equations, and a parallel method looping over squares is proposed in CUDA parallel platform, furthermore. The GPU numerical results for a user-created benchmark structures are checked with comparison to CPU results. A noticeable speedup of the filling of impedance matrix (hundreds times) and solving of linear equations (about 10 times) is achieved due to the employing of GPU.
Keywords :
graphics processing units; impedance matrix; method of moments; parallel architectures; CUDA parallel platform; GPU; NVIDIA company; RWG basis functions; Rao-Wilton-Glisson basis functions; arbitrary configuration; compute unified device architecture; fast electromagnetic simulation; graphics processing unit; impedance matrix elements; linear equations; method of moments; parallel LU decomposition method; parallel MoM; user-created benchmark structures; Equations; Filling; Graphics processing units; Impedance; Mathematical model; Matrix decomposition; Method of moments; CUDA; GPU; LU Decomposition; Method of Moments;
Conference_Titel :
Antennas & Propagation (ISAP), 2013 Proceedings of the International Symposium on
Conference_Location :
Nanjing
Print_ISBN :
978-7-5641-4279-7
