An up to date Mapping Methodology for GPUs
Résumé
—In this paper, we propose an up to date mapping methodology for Graphics Processing Unit (GPU) architectures. The first objective is to analyse and transform the application in order to highlight the minimum parallelism necessary to fit the GPU characteristics. The second is to optimize the mapping of the application onto the GPU by taking into account all its powerful components.
Origine : Fichiers produits par l'(les) auteur(s)
Loading...