Finite Element Analyses of Dynamic Problems Using Graphics Hardware
Atsuya Oishi; and Shinobu Yoshimura

doi:10.3970/cmes.2008.025.115
Source CMES: Computer Modeling in Engineering & Sciences, Vol. 25, No. 2, pp. 115-132, 2008
Download Full length paper in PDF format. Size = 325,792 bytes
Keywords Graphics Hardware, Graphics Processing Unit, Finite Element Method, Dynamic Problem, Explicit Scheme, Element Matrix.
Abstract This paper describes the finite element analyses of dynamic problems using graphics hardware. The graphics hardware, known as GPU that is an acronym of Graphics Processing Unit, was first developed only for processing 3D computer graphics. However it has obtained both flexible programmability using a high-level shader programming language such as OpenGL Shading Language (GLSL), and has also obtained fast numerical processing ability of over 160 GFLOPS that is much faster than CPU. In this paper, GPU is utilized for the finite element analyses of dynamic problems. Two different computational tasks in the dynamic finite element analyses are implemented to the GPU. One is the construction of element stiffness/mass matrices and the other is the calculation of time integration based on an explicit scheme. Fundamental formulations of the implementations are described in detail, and their basic performance is tested through sample analyses. The results indicate that GPU can perform dynamic finite element analyses faster than CPU.
PDF download PDF