Is GPU the future of Scientific Computing ?

Georges-Henri Cottet[1]; Jean-Matthieu Etancelin[1]; Franck Perignon[1]; Christophe Picard[2]; Florian De Vuyst[3]; Christophe Labourdette[3]

  • [1] Laboratoire Jean Kuntzmann Université Joseph Fourier BP 53 38041, Grenoble Cedex 9 France
  • [2] Laboratoire Jean Kuntzamnn Université Joseph Fourier BP 53 38041, Grenoble Cedex 9 France
  • [3] Centre de Mathématiques et de leurs Applications CMLA CNRS UMR 8536 61, avenue du Président Wilson 94235 Cachan CEDEX FRANCE

Annales mathématiques Blaise Pascal (2013)

  • Volume: 20, Issue: 1, page 75-99
  • ISSN: 1259-1734

Abstract

top
These past few years, new types of computational architectures based on graphics processors have emerged. These technologies provide important computational resources at low cost and low energy consumption. Lots of developments have been done around GPU and many tools and libraries are now available to implement efficiently softwares on those architectures.This article contains the two contributions of the mini-symposium about GPU organized by Loïc Gouarin (Laboratoire de Mathématiques d’Orsay), Alexis Hérault (CNAM) and Violaine Louvet (Institut Camille Jordan). This mini-symposium was an opportunity to explore the upcoming role of hardware accelerators and how it will affect the way applications are designed and developed.As the main issue of the mini-symposium was graphical cards, this document contains contributions about two feedbacks on the behavior of different numerical methods on GPU:ones on particle method for transport equations,the other on Lattice Boltzmann Methods for Navier–Stokes equations, Finite Volume schemes for Euler equations and particles methods for kinetic equations.

How to cite

top

Cottet, Georges-Henri, et al. "Is GPU the future of Scientific Computing ?." Annales mathématiques Blaise Pascal 20.1 (2013): 75-99. <http://eudml.org/doc/275429>.

@article{Cottet2013,
abstract = {These past few years, new types of computational architectures based on graphics processors have emerged. These technologies provide important computational resources at low cost and low energy consumption. Lots of developments have been done around GPU and many tools and libraries are now available to implement efficiently softwares on those architectures.This article contains the two contributions of the mini-symposium about GPU organized by Loïc Gouarin (Laboratoire de Mathématiques d’Orsay), Alexis Hérault (CNAM) and Violaine Louvet (Institut Camille Jordan). This mini-symposium was an opportunity to explore the upcoming role of hardware accelerators and how it will affect the way applications are designed and developed.As the main issue of the mini-symposium was graphical cards, this document contains contributions about two feedbacks on the behavior of different numerical methods on GPU:ones on particle method for transport equations,the other on Lattice Boltzmann Methods for Navier–Stokes equations, Finite Volume schemes for Euler equations and particles methods for kinetic equations.},
affiliation = {Laboratoire Jean Kuntzmann Université Joseph Fourier BP 53 38041, Grenoble Cedex 9 France; Laboratoire Jean Kuntzmann Université Joseph Fourier BP 53 38041, Grenoble Cedex 9 France; Laboratoire Jean Kuntzmann Université Joseph Fourier BP 53 38041, Grenoble Cedex 9 France; Laboratoire Jean Kuntzamnn Université Joseph Fourier BP 53 38041, Grenoble Cedex 9 France; Centre de Mathématiques et de leurs Applications CMLA CNRS UMR 8536 61, avenue du Président Wilson 94235 Cachan CEDEX FRANCE; Centre de Mathématiques et de leurs Applications CMLA CNRS UMR 8536 61, avenue du Président Wilson 94235 Cachan CEDEX FRANCE},
author = {Cottet, Georges-Henri, Etancelin, Jean-Matthieu, Perignon, Franck, Picard, Christophe, De Vuyst, Florian, Labourdette, Christophe},
journal = {Annales mathématiques Blaise Pascal},
keywords = {GPU; méthode particulaire; EDP; Mécanique des Fluides; interaction; visualisation; calcul instantané; volumes finis; méthode Lattice Boltzmann; méthode particulaire; programmation multicœur; graphics processing unit},
language = {eng},
month = {1},
number = {1},
pages = {75-99},
publisher = {Annales mathématiques Blaise Pascal},
title = {Is GPU the future of Scientific Computing ?},
url = {http://eudml.org/doc/275429},
volume = {20},
year = {2013},
}

TY - JOUR
AU - Cottet, Georges-Henri
AU - Etancelin, Jean-Matthieu
AU - Perignon, Franck
AU - Picard, Christophe
AU - De Vuyst, Florian
AU - Labourdette, Christophe
TI - Is GPU the future of Scientific Computing ?
JO - Annales mathématiques Blaise Pascal
DA - 2013/1//
PB - Annales mathématiques Blaise Pascal
VL - 20
IS - 1
SP - 75
EP - 99
AB - These past few years, new types of computational architectures based on graphics processors have emerged. These technologies provide important computational resources at low cost and low energy consumption. Lots of developments have been done around GPU and many tools and libraries are now available to implement efficiently softwares on those architectures.This article contains the two contributions of the mini-symposium about GPU organized by Loïc Gouarin (Laboratoire de Mathématiques d’Orsay), Alexis Hérault (CNAM) and Violaine Louvet (Institut Camille Jordan). This mini-symposium was an opportunity to explore the upcoming role of hardware accelerators and how it will affect the way applications are designed and developed.As the main issue of the mini-symposium was graphical cards, this document contains contributions about two feedbacks on the behavior of different numerical methods on GPU:ones on particle method for transport equations,the other on Lattice Boltzmann Methods for Navier–Stokes equations, Finite Volume schemes for Euler equations and particles methods for kinetic equations.
LA - eng
KW - GPU; méthode particulaire; EDP; Mécanique des Fluides; interaction; visualisation; calcul instantané; volumes finis; méthode Lattice Boltzmann; méthode particulaire; programmation multicœur; graphics processing unit
UR - http://eudml.org/doc/275429
ER -

References

top
  1. M. Bergdorf, P. Koumoutsakos, A Lagrangian particle-wavelet method, Multiscale Models. Simul. 5 (2006), 980-995 Zbl1122.65085MR2272307
  2. R.A. Brownlee, A.N. Gorban, J. Levesley, Stabilization of the lattice Boltzmann method using the Ehrenfests’ coarse-graining data, Physical Review E 74 (2006) 
  3. Carlo Cercignani, The Boltzmann Equation and Its Applications, 67 (1988), Springer-Verlag Zbl0646.76001MR1313028
  4. E. Godlewski, P.-A. Raviart, Numerical approximation of hyperbolic conservation laws, 118 (1996), Applied Mathematical Sciences, Springer-Verlag, Boston Zbl0860.65075MR1410987
  5. D. Hanel, R. Schwane, An implicit Flux-Vector Splitting Scheme for the computation of viscous hypersonic flow, AIAA Paper 25 (1989) 
  6. A. Harten, P.D. Lax, B. van Leer, On upstream differencing and Godunov-type schemes for hyperbolic conservation laws, SIAM Review 25 (1983), 35-61 Zbl0565.65051MR693713
  7. A. Magni, G.H. Cottet, Accurate, non-oscillatory, remeshing schemes for particle methods, J. Comput. Phys. 231 (2012), 152-172 Zbl06044227MR2846992
  8. A. Munshi, The OpenCL Specification, Khronos OpenCL Working Group (2011) 
  9. RR. Nourgaliev, T.N. Dinh, T.G. Theofanous, D. Joseph, The lattice Boltzmann equation method: theoretical interpretation, numerics and implications, Int. J. of Multiphase Flow 29 (2003), 117-169 Zbl1136.76594
  10. CUDA C Best Practices Guide 4.1, NVIDIA, (2012) 
  11. D. Rossinelli, M. Bergdorf, G.H. Cottet, P. Koumoutsakos, GPU accelerated simulations of bluff body flows using vortex methods, J. Comput. Phys. 229 (2010), 3316-3333 Zbl1307.76066MR2601102
  12. Sauro Succi, The Lattice Boltzmann Equation for Fluid Dynamics and Beyond, (2001), Oxford Zbl0990.76001MR1857912
  13. F. De Vuyst, A Flux Vector Splitting method that preserves stationary contact discontinuities, Acta Mathematicae Applicandae (2013) 
  14. F. De Vuyst, F. Salvarani, GPU-Accelerated numerical simulations of the Knudsen gas on time-dependent domains, Computer Physics Communications 184 (2013), 532-536 Zbl1302.76157MR3007037

NotesEmbed ?

top

You must be logged in to post comments.

To embed these notes on your page include the following JavaScript code on your page where you want the notes to appear.

Only the controls for the widget will be shown in your chosen language. Notes will be shown in their authored language.

Tells the widget how many notes to show per page. You can cycle through additional notes using the next and previous controls.

    
                

                
Note: Best practice suggests putting the JavaScript code just before the closing </body> tag.