The best paths to the exascale summit are debatable, but all are narrow and treacherous, constrained by fundamental laws of physics, capital cost, operating cost, power requirements, programmability, and reliability. Many scientific and engineering applications force the modeling community to attempt to scale this summit. Drawing on vendor projections and experiences with scientific codes on contemporary platforms, we outline the challenges and propose roles and essential adaptations for mathematical modelers in one of the great global scientific quests the next decade.
On peut discuter du meilleur chemin vers les sommets de l'exascale mais tous sont traitres et étroits, conditionnés par les lois fondamentales de la physique, les couts d'investissement et de fonctionnement, la consommation électrique, la facilité de programmation et la fiabilité. Pour la plus part des applications la science et l'ingénierie, la communauté des modélisateurs sera forcée de s'adapter pour gravir ce sommet. Partant des prévisions des constructeurs et de notre expérience des codes scientifiques sur les plateformes actuelles, nous tracerons les grandes lignes du défit exascale et proposerons des scénarios et des adaptations aux numériciens pour ce grand challenge scientifique de la décennie.
Mots-clés : Informatique, Exaflop
David E. Keyes 1, 2
@article{CRMECA_2011__339_2-3_70_0, author = {David E. Keyes}, title = {Exaflop/s: {The} why and the how}, journal = {Comptes Rendus. M\'ecanique}, pages = {70--77}, publisher = {Elsevier}, volume = {339}, number = {2-3}, year = {2011}, doi = {10.1016/j.crme.2010.11.002}, language = {en}, }
David E. Keyes. Exaflop/s: The why and the how. Comptes Rendus. Mécanique, High Performance Computing, Volume 339 (2011) no. 2-3, pp. 70-77. doi : 10.1016/j.crme.2010.11.002. https://comptes-rendus.academie-sciences.fr/mecanique/articles/10.1016/j.crme.2010.11.002/
[1] Partial differential equation-based applications and solver algorithms at extreme scale, Int. J. High Perf. Comput. Appl., Volume 23 (2009), pp. 366-368
[2] A bridging model for parallel computation, Comm. ACM, Volume 33 (1990), pp. 103-111
[3] High performance parallel implicit CFD, Parallel Comput., Volume 27 (2001), pp. 337-362
[4] A view of the parallel computing landscape, Comm. ACM, Volume 52 (2009), pp. 56-67
[5] J. Dongarra, P. Beckman, et al., The international exascale software project roadmap, Int. J. High Perf. Comput. Appl. 25 (2011), in press.
[6] G. Bell, Great and big ideas in computer structures, in: Mind Matters: A Tribute to Allen Newell, 1996, pp. 189–218.
[7] PETSc Web page, 2010 http://www.mcs.anl.gov/petsc/
[8] Scalable adaptive mantle convection simulation on petascale supercomputers, SC'08: Proceedings of the 2008 ACM/IEEE Conference on Supercomputing, IEEE Computer Society, 2008, pp. 1-15 ([Gordon Bell Prize finalist.])
[9] D.E. Keyes, How scalable is domain decomposition in practice? in: C.-H. Lai, et al. (Eds.) Proceedings of the 11th Intl. Conf. on Domain Decomposition Methods, 1998, pp. 286–297.
[10] Chaotic relaxation, Linear Algebra Appl., Volume 2 (1969), pp. 199-222
[11] Asynchronous fast adaptive composite grid methods: Theoretical foundations, SIAM J. Numer. Anal., Volume 42 (2004), pp. 130-152
[12] T.P. Collignon, M.B. Van Gijzen, Fast iterative solution of large sparse linear systems on geographically separated clusters, Int. J. High Perf. Comput. Appl. (2010), in press.
[13] More scalability, less pain: A simple programming model and its implementation for extreme computing, SciDAC Rev., Volume 17 (2010), pp. 30-37
[14] Communication-optimal iterative methods, J. Phys. Conf. Ser., Volume 180 (2009), p. 012040
[15] D. Brown, P. Messina, P. Beckman, D. Keyes, J. Vetter, Crosscutting technologies for computing at extreme scale, U.S. Department of Energy, 2010, in press.
[16] Letting physicists be physicists, and other goals of scalable solver software (A. Deane, ed.), Proceedings of Parallel CFD'05, Elsevier, 2005, pp. 51-75
[17] J. Dean, S. Ghemawat, MapReduce: Simplified data processing on large clusters, in: OSDI'04: Sixth Symposium on Operating System Design and Implementation, San Francisco, CA, December 2004.
[18] J.F. Kennedy, We choose to go to the Moon. Speech delivered at Rice University, Houston, Texas on 12 September 1962.
Cited by Sources:
Comments - Policy