Identifying scalar behavior in CUDA kernels - Archive ouverte HAL Accéder directement au contenu
Rapport (Rapport De Recherche) Année : 2011

Identifying scalar behavior in CUDA kernels

Caroline Collange
Connectez-vous pour contacter l'auteur

Résumé

We propose a compiler analysis pass for programs expressed in the Single Program, Multiple Data (SPMD) programming model. It identifies statically several kinds of regular patterns that can occur between adjacent threads, including common computations, memory accesses at consecutive locations or at the same location and uniform control flow. This knowledge can be exploited by SPMD compilers targeting SIMD architectures. We present a compiler pass developed within the Ocelot framework that performs this analysis on NVIDIA CUDA programs at the PTX intermediate language level. Results are compared with optima obtained by simulation of several sets of CUDA benchmarks.
Fichier principal
Vignette du fichier
Collange_ScalarizingCompiler_TR11.pdf (189.54 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-00555134 , version 1 (12-01-2011)

Identifiants

  • HAL Id : hal-00555134 , version 1

Citer

Caroline Collange. Identifying scalar behavior in CUDA kernels. [Research Report] ENS Lyon. 2011. ⟨hal-00555134⟩
319 Consultations
750 Téléchargements

Partager

Gmail Facebook X LinkedIn More