TLM OpenCL multi-GPUs implementation - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2012

TLM OpenCL multi-GPUs implementation

Résumé

Time-domain methods attracts much attention since a few years due to their ability to describe sound propagation in complex environments. However, such approaches requires a substantial computational power and a super-computer remains still relevant for sound propagation simulations in huge domains. Graphics processing units (GPUs) are massively parallel computing environments, that are more and more used for scientific computations purposes. Among time-domain methods, the transmission line modeling (TLM) method is especially well adapted to this kind of hardware architecture. The paper presents an OpenCL implementation of the TLM method, this language being compatible with a wide range of available computational hardware, either with Central Processing Units (CPUs) or with GPUs devices. The algorithm is applied to evaluate the impact of various scenarios of vegetation cover on urban frontages and building rooftops within the framework of the VegDUD french project.
Fichier principal
Vignette du fichier
Guillaume_-_TLM_openCL_multi-GPUs_implementation.pdf (1.44 Mo) Télécharger le fichier
Origine : Accord explicite pour ce dépôt
Loading...

Dates et versions

hal-00804170 , version 1 (25-03-2013)

Identifiants

  • HAL Id : hal-00804170 , version 1

Citer

Gwenaël Guillaume, Nicolas Fortin, Benoit Gauvreau, Judicaël Picaut. TLM OpenCL multi-GPUs implementation. Acoustics 2012, Apr 2012, Nantes, France. pp.3249-3255. ⟨hal-00804170⟩
144 Consultations
270 Téléchargements

Partager

Gmail Facebook X LinkedIn More