Reconvergence de contrôle implicite pour les architectures SIMT

Nicolas Brunie 1, 2, 3 Sylvain Collange 4, *
* Corresponding author
2 ARIC - Arithmetic and Computing
Inria Grenoble - Rhône-Alpes, LIP - Laboratoire de l'Informatique du Parallélisme
4 ALF - Amdahl's Law is Forever
Inria Rennes – Bretagne Atlantique , IRISA-D3 - ARCHITECTURE
Abstract : Parallel architectures following the SIMT model such as GPUs benefit from application regularity by issuing concurrent threads running in lockstep on SIMD units. As threads take different paths across the control-flow graph, lockstep execution is partially lost, and must be regained whenever possible in order to maximize the occupancy of SIMD units. In this paper, we propose two techniques to handle SIMT control divergence and identify reconvergence points. The most advanced one operates in constant space and handles indirect jumps and recursion. We evaluate a hardware implementation which leverage the existing memory divergence management unit. In terms of performance, this solution is at least as efficient as state of the art techniques in use in current GPUs.
Document type :
Journal articles
Complete list of metadatas

Cited literature [28 references]  Display  Hide  Download

https://hal.inria.fr/hal-00787749
Contributor : Caroline Collange <>
Submitted on : Tuesday, February 12, 2013 - 6:04:49 PM
Last modification on : Friday, September 20, 2019 - 4:36:06 PM
Long-term archiving on : Saturday, April 1, 2017 - 10:24:15 PM

File

Brunie_ReconvImpliciteSIMT_TR....
Files produced by the author(s)

Identifiers

Citation

Nicolas Brunie, Sylvain Collange. Reconvergence de contrôle implicite pour les architectures SIMT. Revue des Sciences et Technologies de l'Information - Série TSI : Technique et Science Informatiques, Lavoisier, 2013, Architecture des ordinateurs, 32 (2), pp.153-178. ⟨10.3166/TSI.32.153-178⟩. ⟨hal-00787749⟩

Share

Metrics

Record views

770

Files downloads

357