Accéder directement au contenu Accéder directement à la navigation
Article dans une revue

Building a Polyhedral Representation from an Instrumented Execution: Making Dynamic Analyses of Non-Affine Programs Scalable

Abstract : The polyhedral model has been successfully used in production compilers. Nevertheless, only a very restricted class of applications can benefit from it. Recent proposals investigated how runtime information could be used to apply polyhedral optimization on applications that do not statically fit the model. In this work, we go one step further in that direction. We propose the folding-based analysis that, from the output of an instrumented program execution, builds a compact polyhedral representation. It is able to accurately detect affine dependencies, fixed-stride memory accesses and induction variables in programs. It scales to real-life applications, which often include some non-affine dependencies and accesses in otherwise affine code. This is enabled by a safe fine-grain polyhedral over-approximation mechanism. We evaluate our analysis on the entire Rodinia benchmark suite, enabling accurate feedback about potential for complex polyhedral transformations.
Type de document :
Article dans une revue
Liste complète des métadonnées

Littérature citée [46 références]  Voir  Masquer  Télécharger

https://hal.inria.fr/hal-02418987
Contributeur : Manuel Selva <>
Soumis le : jeudi 19 décembre 2019 - 11:30:03
Dernière modification le : jeudi 19 novembre 2020 - 13:02:12
Archivage à long terme le : : vendredi 20 mars 2020 - 16:23:43

Fichier

taco-hal.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

Collections

Citation

Manuel Selva, Fabian Gruber, Diogo Sampaio, Christophe Guillon, Louis-Noël Pouchet, et al.. Building a Polyhedral Representation from an Instrumented Execution: Making Dynamic Analyses of Non-Affine Programs Scalable. ACM Transactions on Architecture and Code Optimization, Association for Computing Machinery, 2019, 16 (4), pp.1-26. ⟨10.1145/3363785⟩. ⟨hal-02418987⟩

Partager

Métriques

Consultations de la notice

89

Téléchargements de fichiers

520