Implementation of selected numerical algorithms for solving sparse matrixes using CUDA technology

Author

Zimon, J. ; Zoworka, M.

Author_Institution

Dept. of Ind. Electr. Eng., Opole Univ. of Technol., Opole, Poland

fYear

2013

fDate

15-18 May 2013

Firstpage

77

Lastpage

78

Abstract

This thesis presents an implementation of Finite Element Method (FEM), with focus on sparse matrix solving methods using NVIDIA CUDA devices. Development of an application GPU_FEM that implements Finite Element Method has been presented. The calculation results and calculation times have been compared for different hardware configuration as well as for different algorithms. Presented tests show up to 70 times faster sparse matrix solving in comparison to FEMM application.

Keywords

finite element analysis; graphics processing units; mathematics computing; parallel architectures; sparse matrices; CUDA technology; FEMM application; GPU_FEM application; NVIDIA CUDA device; finite element method; hardware configuration; numerical algorithm; sparse matrix; Equations; Finite element analysis; Graphics processing units; Jacobian matrices; Libraries; Mathematical model; Sparse matrices; Finite Element Method; NVIDIA CUDA technology; Numerical calculations; multithread programming; sparse matrixes;

fLanguage

English

Publisher

ieee

Conference_Titel

Electrodynamic and Mechatronic System (SELM), 2013 International Symposium on

Conference_Location

Opole-Zawiercie

Print_ISBN

978-1-4673-5588-9

Type

conf

DOI

10.1109/SELM.2013.6562988

Filename

6562988