Parallel Computing of Central Difference Explicit Finite Element Based on GPU General Computing Platform

Explicit finite element method has been widely used for the plane nonlinear dynamic problems. Because of the limitation of time step due to the conditional stability, the analysis of large-scale problem always requires long computing time. Graphics processing unit (GPU) is a parallel device with single instruction, multiple data classification. GPU offers high computation power and increases memory bandwidth at a relatively low cost, and it is well suited for problems that can be expressed as data-parallel computations with high arithmetic intensity. Nowadays, it has gained more and more attention as a kind of general parallel processor, followed by various general purpose GPU computing technologies represented by NVIDIA CUDA. In this paper, a method of explicit finite element parallel computing based on central difference method and GPU general computing platform for plane nonlinear dynamic problems is developed. The original serial algorithm is adjusted and optimized for GPU computing based on the characteristics of GPU. Finally, GPU is used for the whole explicit iterative process by mapping one element or one node to one CUDA thread, and the iterative process can be solved parallelly in any order. The numerical examples indicate that this method can greatly improve the computational efficiency with the same computing precision on the NVIDIA GTX 460, and it provides an efficient and simple method for the plane nonlinear dynamic problem.