GPU Cuda Performance on Two-Dimensional and Three-Dimensional VAWT Vortex Models
Mechanical and Aerospace Department, West Virginia University, Morgantown, 26506, United States
International Journal of Scientific and Statistical Computing (IJSSC), Vol. 2, Issue 1, 2011
@article{hard2011gpu,
title={GPU Cuda Performance on Two-Dimensional and Three-Dimensional VAWT Vortex Models},
author={Hard, S. and Wilhelm, J. and Pertl, F.A. and Smith, J.E.},
year={2011}
}
An analytical model of a vertical axis wind turbine was implemented using both a 2-D and a 3-D vortex model. The model requires significant amounts of computational resources and time compelling the use of an improved method for executing the algorithm in a highly parallelized fashion. Graphics Processing Units (GPUs), which are a new highly cost effective method of implementing parallel processing, were used along with a interface code to allow the numerically intensive routine to execute as parallel code on a graphics card. This achieved a substantially reduced computation time compared to a traditional desktop computer model. Using two quad-core processors and a graphics card, a 19.7 fold reduction in computation time was achieved. This paper will describe the model used, the parallel code created, the GPU interface, and an illustration of the performance increases.
November 10, 2011 by hgpu
