时域有限差分法中的GPU加速高效CPML方案

doi:10.3969/j.issn.1001-2400.2015.01.031

J4 ›› 2015, Vol. 42 ›› Issue (1): 194-199+212.doi: 10.3969/j.issn.1001-2400.2015.01.031

时域有限差分法中的GPU加速高效CPML方案

白冰;牛中奇

(西安电子科技大学电子工程学院，陕西西安 710071)

收稿日期:2014-06-06 出版日期:2015-02-20 发布日期:2015-04-14
通讯作者: 白冰
作者简介:白冰(1982-)，男，西安电子科技大学博士研究生，E-mail：bbai@mail.xidian.edu.cn．
基金资助:
国家自然科学基金资助项目(30870577， 61301288)；中央高校基本科研业务费资助项目(JB140218，K5051302057)

High performance CPML acceleration scheme with GPU for FDTD

BAI Bing;NIU Zhongqi

(School of Electronic Engineering, Xidian Univ., Xi'an 710071, China)

Received:2014-06-06 Online:2015-02-20 Published:2015-04-14
Contact: BAI Bing

摘要/Abstract

摘要：

针对并行CPML存在的计算冗余和访问冗余问题，提出了一种用于时域有限差分法的图形处理器加速无除法联合最小访存CPML更新方案．该方案通过重新安排CPML迭代公式，将除法操作吸收进公式的固定系数中，消去了图形处理器计算中负担繁重的除法操作．该方案进一步通过合并PML区域内时域有限差分法常规场值更新步骤和CPML更新步骤，剔除了这两个步骤中的重复访存，使算法的访存需求最小化．数值验证结果表明，在同等精度下，CPML更新过程和PML区域场值整体计算过程分别减少了70%和44%的计算时间．

关键词: 时域有限差分法, 卷积完全匹配层, 图形处理器, 并行计算, 计算统一设备架构

Abstract:

To overcome computational redundancy and memory-access redundancy of the traditional GPU-accelerated CPML technique, a novel division-free and minimum-access CPML scheme is proposed. In the proposed scheme, the division operators in the CPML method are merged into a series of fixed coefficients by optimally rearranging the iteration process of CPML and then, the reduplicate memory accesses are eliminated by updating the FDTD and CPML operation in the PML region jointly. Experimental results show that the proposed structure can save up to 70% operation time compared with the traditional GPU-CPML technique and 44% of field updating in the PML region, without any loss of accuracy.

Key words: finite difference time domain method, convolution perfectly matched layer, graphics processing unit, parallel computing, compute unified device architecture

中图分类号:

TP391.9

白冰;牛中奇. 时域有限差分法中的GPU加速高效CPML方案[J]. J4, 2015, 42(1): 194-199+212.

BAI Bing;NIU Zhongqi. High performance CPML acceleration scheme with GPU for FDTD[J]. J4, 2015, 42(1): 194-199+212.

参考文献

［1］ Taflove A, Hagness S C. Computational Electrodynamics: the Finite Difference Time Domain Method ［M］. 3rd Edition. Norwood: Artech House Publishers, 2005: 273-328.
［2］姜彦南, 葛德彪, 杨利霞, 等. 二维半空间时域有限差分瞬态场外推方法［J］. 西安电子科技大学学报, 2014, 41(2): 178-184.
Jiang Yannan, Ge Debiao, Yang Lixia, et al. Extrapolation to the Transient Ffield for FDTD in Ttwo-dimensional Hhalf-space［J］. Journal of Xidian University, 2014, 41(2): 178-184.
［3］张连波, 郭立新, 苟雪银, 等. 三层粗糙面电磁散射的矩量法研究［J］. 西安电子科技大学学报,2013, 40(6): 147-154.
Zhang Lianbo, Guo Lixin, Gou Xueyin, et al. Method of Moment Investigation on Electromagnetic Scattering from the Three-layered Rough Interfaces［J］. Journal of Xidian University, 2013, 40(6): 147-154.
［4］ Berenger J P. A Perfectly Matched Layer for the Absorption of Electromagnetic Waves［J］. Journal of Computational Physics, 1994, 114(2): 185-200.
［5］ Kuzuoglu M, Mittra R. Frequency Dependence of the Constitutive Parameters of Causal Perfectly Matched Anisotropic Absorbers ［J］. Microwave and Guided Wave Letters, 1996, 6(12): 447-449.
［6］ Roden J, Gedney S D. Convolution PML(CPML): an Efficient FDTD Implementation of the CFS-PML for Arbitrary Media ［J］. Microwave and Optical Technology Letters, 2000, 27(5): 334-339.
［7］ Zygiridis T T, Kantartzis N V, Tsiboukis T D. GPU-Accelerated Efficient Implementation of FDTD Methods with Optimum Time-step Selection［J］. IEEE Transactions on Magnetics, 2014, 50(2): 477-480.
［8］ Sypek P, Dziekonski A, Mrozowski M. How to Render FDTD Computations More Effective Using a Graphics Accelerator［J］. IEEE Transactions on Magnetics, 2009, 45(3): 1324-1327.
［9］ Inman M J, Elsherbeni A Z, Maloney J G, et al. GPU Based FDTD Solver with CPML Boundaries［C］//Antennas and Propagation Society International Symposium. Piscataway: IEEE, 2007: 5255-5258.
［10］胡媛, 李康, 孔凡敏, 等. 基于CUDA架构的三维CPML-FDTD并行方法［J］. 计算机工程与应用, 2011, 47(25): 220-223.
Hu Yuan, Li Kang, Kong Fanmin, et al. Three Dimensional CPML-FDTD Parallel Algorithm Based on CUDA［J］. Computer Engineering and Applications, 2011, 47(25): 220-223.
［11］ Chen Guanbo, Moghaddam M. GPU Accelerated 3D Nonlinear Time Domain Inversion of Realistic Breast Phantoms with Multiparameter Optimization［C］//Radio Science Meeting(Joint with AP-S Symposium). Piscataway: IEEE, 2013: 206.

[1]	翟畅,林中朝,赵勋旺,张玉. 一种使用八叉树的半空间MLFMA区域分解算法[J]. 西安电子科技大学学报, 2021, 48(6): 144-150.
[2]	张玉强;葛德彪. 一种高稳定的复杂色散介质分析FDTD方法[J]. 西安电子科技大学学报, 2018, 45(4): 118-122+173.
[3]	姜雪松;江树刚;张玉;赵勋旺. 国产众核处理器百万核时域有限差分并行计算[J]. 西安电子科技大学学报, 2017, 44(6): 65-69+128.
[4]	任计林;车书玲;郑征. 规则LDPC码在GPU上的加速译码[J]. 西安电子科技大学学报, 2017, 44(3): 25-30.
[5]	江树刚;林中朝;张玉;魏兵;赵勋旺. 国产超级计算机实现10万核FDTD并行计算[J]. J4, 2015, 42(5): 86-91.
[6]	周航;蔡志明;王希敏. 宽带信号匹配滤波的GPU实现及性能优化[J]. J4, 2015, 42(3): 135-140+191.
[7]	刘镇弢;李涛;黄虎才;韩俊刚;沈绪榜 . 一种用于实时图像处理的众核结构设计[J]. J4, 2015, 42(2): 95-101.
[8]	武勇;王俊;张培川;曹运合. CUDA架构下外辐射源雷达杂波抑制并行算法[J]. J4, 2015, 42(1): 104-111.
[9]	秦华;周沫;察豪;左炜. 软件雷达信号处理的多GPU并行技术[J]. J4, 2013, 40(3): 145-151.
[10]	雷继兆;梁昌洪;张玉. 并行FDTD结合服务器分析电大电磁问题[J]. J4, 2009, 36(5): 846-850.
[11]	丁海1;褚庆昕2. TE模式下子域二阶精度FDTD算法 [J]. J4, 2009, 36(1): 162-165.
[12]	陈智慧1;褚庆昕2. 任意线性集总网络FDTD建模新方法 [J]. J4, 2008, 35(2): 262-266.
[13]	孙宏元1;2;谢维信2;陆克中2;杨勋1. 一种新的视界覆盖遗传算法 [J]. J4, 2007, 34(5): 758-762.
[14]	张鹏飞;龚书喜. 基于CRE和MPI的腔体RCS快速并行计算 [J]. J4, 2007, 34(1): 82-86.
[15]	王林年;褚庆昕;梁昌洪. 交替方向隐式时域有限差分法中的Berenger理想匹配层[J]. J4, 2006, 33(3): 458-461.

时域有限差分法中的GPU加速高效CPML方案

High performance CPML acceleration scheme with GPU for FDTD

PDF (PC)

赞

可视化

被引次数

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

Metrics

本文评价

推荐阅读 10