High Performance by Exploiting Information Locality through Reverse Computing - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Reports (Research Report) Year : 2011

High Performance by Exploiting Information Locality through Reverse Computing

Abstract

In this paper we present performance results for our register rematerialization technique based on reverse recomputing. Rematerialization adds instructions and we show on one specifically designed example that reverse computing alleviates the impact of these additional instructions on performance. We also show how thread parallelism may be optimized on GPUs by performing register allocation with reverse recomputing that increases the number of threads per Streaming Multiprocessor (SM). This is done on the main kernel of Lattice Quantum ChromoDynamics (LQCD) simulation program where we gain a 10.84% speedup.
Fichier principal
Vignette du fichier
bahi_information_locality.pdf (1.04 Mo) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

inria-00615493 , version 1 (19-08-2011)

Identifiers

  • HAL Id : inria-00615493 , version 1

Cite

Mouad Bahi, Christine Eisenbeis. High Performance by Exploiting Information Locality through Reverse Computing. [Research Report] 2011. ⟨inria-00615493⟩
133 View
167 Download

Share

Gmail Facebook Twitter LinkedIn More