基于特征基函数及并行技术的高性能算法研究
发布时间：2018-10-29 09:34
【摘要】：基于矩量法的快速计算方法是计算电磁学的一个研究热点,而特征基函数方法是其中一种快速方法。本文研究了宽频带特征基函数方法,并基于多核CPU和GPU并行平台设计了其并行实现。本文主要工作如下:(1)提出了一种快速求解宽频带电磁散射特性的插值型宽频带特征基函数方法。该方法在特征基函数方法的基础上,通过插值形成阻抗矩阵和全域特征基函数,有效减少了未知量的个数,降低了对数据存储的要求,实现了对宽频带电磁散射问题的快速求解。(2)提出了一种基于OpenMP的多核CPU并行加速方法。该方法采用基于频点的并行加速方案,充分发挥多核CPU的计算能力,实现了插值型宽频带特征基函数方法基于多核CPU并行平台的加速计算。(3)提出了一种基于CUDA编程模型快速填充矩阵的并行加速方法。该方法利用CUDA编程模型的两层并行机制,能够对矩阵乘积进行快速计算,加速了约化矩阵的填充,实现了插值型宽频带特征基函数方法基于GPU平台的并行加速计算。
[Abstract]:The fast calculation method based on the method of moments is a hot topic in computational electromagnetics, and the feature basis function method is one of the fast methods. In this paper, the broadband feature basis function method is studied, and its parallel implementation is designed based on multi-core CPU and GPU parallel platform. The main work of this paper is as follows: (1) an interpolated broadband eigenfunction method is proposed to solve the wide-band electromagnetic scattering characteristics quickly. Based on the feature basis function method, the impedance matrix and the global feature basis function are formed by interpolation, which can effectively reduce the number of unknown variables and reduce the requirement of data storage. The fast solution of wide-band electromagnetic scattering problem is realized. (2) A multi-core CPU parallel acceleration method based on OpenMP is proposed. In this method, the parallel acceleration scheme based on frequency points is used to give full play to the computing power of multi-core CPU. The interpolation broadband feature basis function method based on multi-core CPU parallel platform is implemented. (3) A parallel acceleration method based on CUDA programming model is proposed. Using the two-layer parallel mechanism of the CUDA programming model, the method can calculate the matrix product quickly, accelerate the filling of the reduced matrix, and realize the parallel accelerated computation based on the GPU platform based on the interpolation broadband feature basis function method.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O441
本文编号：2297322
[Abstract]:The fast calculation method based on the method of moments is a hot topic in computational electromagnetics, and the feature basis function method is one of the fast methods. In this paper, the broadband feature basis function method is studied, and its parallel implementation is designed based on multi-core CPU and GPU parallel platform. The main work of this paper is as follows: (1) an interpolated broadband eigenfunction method is proposed to solve the wide-band electromagnetic scattering characteristics quickly. Based on the feature basis function method, the impedance matrix and the global feature basis function are formed by interpolation, which can effectively reduce the number of unknown variables and reduce the requirement of data storage. The fast solution of wide-band electromagnetic scattering problem is realized. (2) A multi-core CPU parallel acceleration method based on OpenMP is proposed. In this method, the parallel acceleration scheme based on frequency points is used to give full play to the computing power of multi-core CPU. The interpolation broadband feature basis function method based on multi-core CPU parallel platform is implemented. (3) A parallel acceleration method based on CUDA programming model is proposed. Using the two-layer parallel mechanism of the CUDA programming model, the method can calculate the matrix product quickly, accelerate the filling of the reduced matrix, and realize the parallel accelerated computation based on the GPU platform based on the interpolation broadband feature basis function method.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O441
【参考文献】
相关博士学位论文 前1条
1 孔维宾;电磁场多尺度问题的混合快速算法研究[D];东南大学;2015年
相关硕士学位论文 前4条
1 常伟;基于并行计算技术的粗糙面与目标电磁散射快速算法研究[D];西安电子科技大学;2014年
2 柴豆豆;基于GPU加速的矩量法对电磁散射问题的研究[D];安徽大学;2012年
3 游佐勇;OpenMP并行编程模型与性能优化方法的研究及应用[D];成都理工大学;2011年
4 麻军;矩量法及其并行计算在粗糙面和目标电磁散射中的应用[D];西安电子科技大学;2009年
，本文编号：2297322
本文链接：https://www.wllwen.com/shoufeilunwen/benkebiyelunwen/2297322.html
