A strategy for array management in local memory

Abstract : One major point in loop restructuring for data locality optimization is the choice and the evaluation of a data locality criteria. We show in this paper how to compute approximations of window sets defined by Gannon, Jalby and Gallivan (the window associated with an iteration i describes the "active" portion of array : elements which have already been referenced before iteration i and which will be referenced after iteration i. Such a notion is extremely useful for data localization since it identifies the portions of arrays which are worth keeping in local memory because they are going to be referenced later. The computation of these window approximations can be performed symbolically at compile time and generates simple geometrical shape that simplifies the management of the data transfers. This allow to derive a global strategy of data management for local memories. . . Moreover, the effects of loop transformations fit naturally in the geometrical framework we use for the calculations. The determination of window approximations is studied both from a theoretical and a computational point of view and examples of applications are given.
Type de document :
Rapport
RR-1262, INRIA. 1990
Liste complète des métadonnées

https://hal.inria.fr/inria-00075296
Contributeur : Rapport de Recherche Inria <>
Soumis le : mercredi 24 mai 2006 - 17:53:41
Dernière modification le : jeudi 26 octobre 2017 - 15:04:02
Document(s) archivé(s) le : dimanche 4 avril 2010 - 21:54:30

Fichiers

Identifiants

  • HAL Id : inria-00075296, version 1

Collections

Citation

Christine Eisenbeis, W. Jalby, D. Windheiser, François Bodin. A strategy for array management in local memory. RR-1262, INRIA. 1990. 〈inria-00075296〉

Partager

Métriques

Consultations de la notice

154

Téléchargements de fichiers

161