Parallel Implementation of a Balancing Domain Decomposition Method for Multi-core Processors

Author

Ogino, Masao ; Kawai, Hiroshi ; Shioya, Ryuji ; Yoshimura, Shinobu

Author_Institution

Fac. of Eng., Kyushu Univ., Fukuoka, Japan

fYear

2010

fDate

4-6 Nov. 2010

Firstpage

56

Lastpage

61

Abstract

To improve the total performance of a PC cluster or a supercomputer with multi-core scalar CPU, not only parallel efficiency among computational nodes or CPUs, but also efficient utilization of processor cores of CPU are becoming more important than ever before. In this paper, we implement the BDD method by three kinds of programming styles to evaluate parallel performance on a multi-core CPU node. We demonstrate some numerical examples using a multi-core CPU. Especially, by changing the design from memory access intensive approach to matrix storage-free one, the parallel performance is improved.

Keywords

finite element analysis; microprocessor chips; BDD method; PC cluster performance; balancing domain decomposition method; memory access intensive approach; multicore processors; multicore scalar CPU; supercomputer; Boolean functions; Central Processing Unit; Data structures; Electronic mail; Iterative methods; Matrix decomposition; Multicore processing; balancing domain decomposition; multi-core processor; parallel computing;

fLanguage

English

Publisher

ieee

Conference_Titel

Broadband, Wireless Computing, Communication and Applications (BWCCA), 2010 International Conference on

Conference_Location

Fukuoka

Print_ISBN

978-1-4244-8448-5

Electronic_ISBN

978-0-7695-4236-2

Type

conf

DOI

10.1109/BWCCA.2010.48

Filename

5633713