UPML-FDTD parallel computing on GPU

Author

Weiwei, Ma ; Dong, Sun ; Xianliang, Wu

Author_Institution

Key of Intell. Comput. & Signal Process. of Minist. of Educ., Anhui Univ., Hefei, China

Volume

4

fYear

2012

fDate

5-8 May 2012

Firstpage

1

Lastpage

4

Abstract

Finite-Difference Time-Domain method (FDTD) which is known with its simple and flexible is widely used for the calculating of electromagnetic fields. However, it costs a lot of time in simulating the electrical-large object. To solve the problem, parallel method for UPML-FDTD algorithm was put forward using GPU based-on Compute Unified Device Architecture (CUDA) in this paper. The algorithm is further optimized by using texture memory. Comparing with the calculation with traditional CPU, The result of simulation shows that this algorithm has enough precision and the remarkable increased efficiency was acquired in different sizes of Yee cells.

Keywords

electromagnetic fields; finite difference time-domain analysis; graphics processing units; CPU; CUDA; GPU based-on compute unified device architecture; UPML-FDTD parallel computing; Yee cells; electrical-large object simulation; electromagnetic fields calculation; finite-difference time-domain method; parallel method; texture memory; Algorithm design and analysis; Computational modeling; Computer architecture; Finite difference methods; Graphics processing unit; Instruction sets; Time domain analysis;

fLanguage

English

Publisher

ieee

Conference_Titel

Microwave and Millimeter Wave Technology (ICMMT), 2012 International Conference on

Conference_Location

Shenzhen

Print_ISBN

978-1-4673-2184-6

Type

conf

DOI

10.1109/ICMMT.2012.6230262

Filename

6230262