Scheduling paths leveraging dynamic information in SIMT architectures - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2021

Scheduling paths leveraging dynamic information in SIMT architectures


Thread divergence optimization in GPU architectures have long been hindered by restrictive control-flow mechanisms based on stacks of execution masks. However, GPU architectures recently began implementing more flexible hardware mechanisms, presumably based on path tables. We leverage this opportunity by proposing a hardware implementation of iteration shifting, a divergence optimization that enables lockstep execution across arbitrary iterations of a loop. Although software implementations of iteration shifting have been previously proposed, implementing this scheduling technique in hardware lets us leverage dynamic information such as divergence patterns and memory stalls. Evaluation using simulation suggest that the expected performance improvements will remain modest or even nonexistent unless the organization of the memory access path is also revisited.
Fichier principal
Vignette du fichier
las_compass2021.pdf (114.84 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-03269966 , version 1 (24-06-2021)


  • HAL Id : hal-03269966 , version 1


Lily Blanleuil, Caroline Collange. Scheduling paths leveraging dynamic information in SIMT architectures. COMPAS 2021 - Conférence francophone d'informatique en Parallélisme, Architecture et Système, Jul 2021, Lyon / Virtual, France. pp.1-6. ⟨hal-03269966⟩
120 View
187 Download


Gmail Facebook X LinkedIn More