GPU Computing and CUDA technology used to accelerate a mesh generator application
Instituto de Ciencias, Universidad Nacional de General Sarmiento, San Miguel, Buenos Aires, Argentina
The 2012 International Conference on Parallel and Distributed Processing Techniques and Applications (PDPTA’12), 2012
@article{gaudiani2012gpu,
title={GPU Computing and CUDA technology used to accelerate a mesh generator application},
author={Gaudiani, A. and Montiel, S. and Pim{‘a}s, J.},
year={2012}
}
The potential of GPU computing used in general purpose parallel programming has been amply shown. These massively parallel many-core multiprocessors are available to any users in every PCs, notebook, game console or workstation. In this work, we present the parallel version of a mesh-generating algorithm and its execution time reduction by using off-the-shelf GPU technology. We use commodities GPUs as a useful CPU co-processor to improve this kind of applications, characterized by a high level of data parallelism. Compared to the sequential algorithm, our techniques achieve 6X overall performance for GPU-CPU implementation; furthermore we achieve 50X speedup when implementing core operations of the algorithm. Results show that GPU provides a helpful platform for high performance computing to improve the execution time of these applications.
September 7, 2012 by hgpu
