European Option Princing on a GPU Cluster

Lokman A. Abbas-Turki (1) , Stéphane Vialle (2, 3)

1 CERMICS - Centre d'Enseignement et de Recherche en Mathématiques et Calcul Scientifique
2 SUPELEC-Campus Metz
3 ALGORILLE - Algorithms for the Grid

The aim of this presentation is a comparison in terms of speed and energy consumption between CPU and GPU clusters using financial application as a benchmark. After a fast introduction on the field of application we will give details on the hardware and software architectures used. Then we will introduce a multiparadigm parallel algorithm, mixing coarse and fine grained parallelism, and its implementations using MPI+OpenMP on CPU cluster and MPI+CUDA on GPU cluster. Finally, some computing and energetic performances on different clusters will be compared.

Domains: Distributed, Parallel, and Cluster Computing [cs.DC]

Lokman A. Abbas-Turki, Stéphane Vialle. European Option Princing on a GPU Cluster. 2009. ⟨hal-00364242⟩