Numerical Parallel Processing Based on GPU with CUDA Architecture

Author

Zou, Chengming ; Xia, Chunfen ; Zhao, Guanghui

Author_Institution

Coll. of Comput. Sci. & Technol., Wuhan Univ. of Technol., Wuhan, China

fYear

2009

fDate

28-29 Dec. 2009

Firstpage

93

Lastpage

96

Abstract

The characteristics of modern graphics processing unit (GPU) is programmable, high price / performance ratio and high speed . It has a strong ability to adapt the parallel calculation, Based on this, the article study the general method of GPU calculating and use compute unified device architecture (CUDA) to design new parallel algorithm to accelerate the matrix inversion and binarization algorithm. The results show that with the increase of matrix dimension, GPU performs much better than CPU in increase multiple.

Keywords

coprocessors; matrix inversion; parallel algorithms; parallel architectures; binarization algorithm; compute unified device architecture; graphics processing unit; matrix inversion; numerical parallel processing; parallel algorithm; parallel calculation; Acceleration; Bandwidth; Central Processing Unit; Computer architecture; Concurrent computing; Graphics; Multicore processing; Parallel algorithms; Parallel processing; Yarn; Binarization algorithm; CUDA; GPU; matrix inversion; parallel algorithm;

fLanguage

English

Publisher

ieee

Conference_Titel

Wireless Networks and Information Systems, 2009. WNIS '09. International Conference on

Conference_Location

Shanghai

Print_ISBN

978-0-7695-3901-0

Electronic_ISBN

978-1-4244-5400-6

Type

conf

DOI

10.1109/WNIS.2009.46

Filename

5381567