Skip to Main content Skip to Navigation
Reports

A Conservative Approach to Handle Full Functions in the Polyhedral Model

Mohamed-Walid Benabderrahmane 1 Cédric Bastoul 1 Louis-Noël Pouchet 1 Albert Cohen 1
1 ALCHEMY - Architectures, Languages and Compilers to Harness the End of Moore Years
LRI - Laboratoire de Recherche en Informatique, UP11 - Université Paris-Sud - Paris 11, Inria Saclay - Ile de France, CNRS - Centre National de la Recherche Scientifique : UMR8623
Abstract : The Polyhedral Model is one of the most powerful framework for automatic optimization and parallelization of high-level programs. It is based on an algebraic representation of program parts and allows to achieve exact data dependence analysis and to apply complex sequences of optimizations seamlessly. After more than twenty years of research and development, this model is now quite mature and reaches production compilers as GCC 4.4 and its GRAPHITE framework. The main limitation of the Polyhedral Model is known to be its restricted application domain. Traditionally, it is used to manipulate very regular program parts only. The goal of this paper is to show this limitation is mostly artificial. We identify the main problem as the code generation step. We propose an extension to the polyhedral representation and to a code generation algorithm that allows to manipulate full functions in the Polyhedral Model.
Complete list of metadata

Cited literature [28 references]  Display  Hide  Download

https://hal.inria.fr/inria-00356818
Contributor : Cédric Bastoul <>
Submitted on : Wednesday, January 28, 2009 - 5:16:08 PM
Last modification on : Thursday, July 8, 2021 - 3:48:10 AM
Long-term archiving on: : Friday, October 12, 2012 - 10:21:03 AM

File

RR-6814.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : inria-00356818, version 1

Collections

Citation

Mohamed-Walid Benabderrahmane, Cédric Bastoul, Louis-Noël Pouchet, Albert Cohen. A Conservative Approach to Handle Full Functions in the Polyhedral Model. [Research Report] RR-6814, INRIA. 2008, pp.16. ⟨inria-00356818⟩

Share

Metrics

Record views

557

Files downloads

219