Semi-Lagrangian implementation method using GPU

Author

Gomez, Yensy ; Hincapie, Juan ; Osorio, John ; Vargas, J.

fYear

2014

fDate

3-5 Sept. 2014

Firstpage

240

Lastpage

248

Abstract

Nowadays plasma has become of great importance in both scientific and industrial areas since nanostructures, used for creating small computational processing devices, are produced by pulse lasers. The Simulation of phenomena such as plasma consumes a big amount of CPUs execution time. This paper intends to suggest a parallel solution of the Semi-Lagrangian using GPUs to solve the Vlasov-Poisson equations in 1D, as a different simulation alternative to PIC (Particle in Cell). To make this process a CUDA was used as a programing framework and an nVidia GTX 580 graphic card as a parallel processing element. The result of the parallelization process allowed accelerating the execution time of 3.7 times faster in comparison to the sequential algorithm.

Keywords

Poisson equation; Vlasov equation; graphics processing units; parallel architectures; parallel programming; physics computing; plasma simulation; plasma transport processes; CUDA; GPU; PIC; Vlasov-Poisson equations; nVidia GTX 580 graphic card; parallel processing element; parallel solution; parallelization process; particle in cell; plasma phenomena simulation; programing framework; semiLagrangian implementation method; Graphics processing units; Kernel; Random access memory; GPUs; HPC; Semi-Lagrangian Method; Vlasov-Poisson;

fLanguage

English

Publisher

ieee

Conference_Titel

Computing Colombian Conference (9CCC), 2014 9th

Conference_Location

Pereira

Type

conf

DOI

10.1109/ColumbianCC.2014.6955340

Filename

6955340