Skip to Main content Skip to Navigation

Exécution efficace de systèmes multi-agents sur GPU

Abstract : These last years have seen the emergence of parallelism in many fields ofcomputer science. This is explained by the stagnation of the frequency ofexecution units at the hardware level and by the increasing usage ofparallel platforms at the software level. A form of parallelism ispresent in multi-agent systems, that facilitate the description ofcomplex systems as a collection of interacting entities. If thesimilarity between this software and this logical parallelism seemsobvious, the parallelization process remains difficult in this casebecause of the numerous dependencies encountered in many multi-agentsystems. In this thesis, we propose a common solution to facilitate theadaptation of these models on a parallel platform such as GPUs. Ourlibrary, MCMAS, provides access to two programming interface tofacilitate this adaptation: a low-level layer providing direct access toOpenCL, MCM, and a high-level set of plugins not requiring any GPU-related knowledge. We study the usage of this library on three existingmulti-agent models : predator-prey, MIOR and Collembola. To prove theinterest of the approach we present a performance study for each modeland an analysis of the various factors contributing to an efficientexecution on GPUs. We finally conclude on a overview of the work andresults presented in the report and suggest future directions to enhanceour solution.
Complete list of metadata

Cited literature [95 references]  Display  Hide  Download
Contributor : Jean-Michel Caricand Connect in order to contact the contributor
Submitted on : Tuesday, November 25, 2014 - 3:48:24 PM
Last modification on : Thursday, January 13, 2022 - 12:00:20 PM
Long-term archiving on: : Friday, April 14, 2017 - 8:40:22 PM


  • HAL Id : tel-01087226, version 1


Guillaume Laville. Exécution efficace de systèmes multi-agents sur GPU. Calcul parallèle, distribué et partagé [cs.DC]. Université de Franche-Comté, 2014. Français. ⟨tel-01087226⟩



Record views


Files downloads