基于GPU加速的高阶矩量法研究与应用  

Research and Application of the High-Order Method of Moments Based on GPU Acceleration

作　　者：马韬[1,2] 陈明生[1] 吴先良[1,2] 刘艺[1,2] 齐琪[1,2] 

机构地区：[1]合肥师范学院电子信息工程学院,合肥230061 [2]安徽大学计算智能与信号处理教育部重点实验室,合肥230039

出　　处：《微波学报》2013年第4期34-37,共4页Journal of Microwaves

基　　金：国家自然科学基金(61001033);教育部科学技术重点项目(212081);安徽高校省级自然科学研究重点项目(KJ2011A240)资助课题

摘　　要：矩量法(MOM)在求解电磁场散射问题时,当未知量数目比较大时,其内存占用和计算时间非常大。基于最佳一致逼近理论构造了高阶矩量法,并引入了计算统一设备架构(CUDA)技术,在图形处理器(GPU)上实现了并行加速计算二维电磁散射问题。实例结果表明,在与快速多极子算法(FMM)相对比下,该方法在较低剖分的情况下,具有很高的计算精度,并且在阻抗矩阵填充和矩矢相乘时的速度大大提升,适用于电大尺寸目标的散射问题。The method of moments （MOM） is usually used to solve the electromagnetic scattering problems. When the number of unknowns is increasing, the memory and computation time consumed becomes insufferable. In this paper, the best uniform approximation theory is used to construct a high-order method of moments, and the Compute Unified Device Ar- chitecture （CUDA） technology based on graphics processor unit （GPU） is introduced to achieve parallel speed up two-di- mensional electromagnetic scattering problems. Numerical results show that compared with the FMM, the proposed method has higher accuracy with lower discretization, and the speed of impedance matrix filling and moment vector multiply is greatly improved, which can be used as an efficient tool to solving the scattering problems of the electrically large targets.

关 键 词：矩量法(MOM) 最佳一致逼近 高阶矩量法 图形处理器(GPU) 计算统一设备架构(CUDA) 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] O441.4[自动化与计算机技术—计算机科学与技术]