PRISM - Parallélisme, Réseaux, Systèmes, Modélisation, UVSQ - Université de Versailles Saint-Quentin-en-Yvelines, Inria Paris-Rocquencourt, CNRS - Centre National de la Recherche Scientifique : UMR8144
Abstract : Program transformations are one of the most valuable compiler techniques to improve data locality. However, restructuring compilers have a hard time coping with data dependences. A typical solution is to focus on program parts where the dependences are simple enough to enable any transformation. For more complex problems is only addressed the question of checking whether a transformation is legal or not. In this paper we propose to go further. Starting from a transformation with no guarantee on legality, we show how we can correct it for dependence satisfaction with no consequence on its locality properties. Generating code having the best locality is a direct application of this result.
https://hal.inria.fr/inria-00001056 Contributor : Cédric BastoulConnect in order to contact the contributor Submitted on : Wednesday, January 25, 2006 - 12:52:32 AM Last modification on : Friday, January 21, 2022 - 3:15:46 AM Long-term archiving on: : Saturday, April 3, 2010 - 9:44:47 PM
Cédric Bastoul, Paul Feautrier. More Legal Transformations for Locality. Euro-Par 2004 Parallel Processing: 10th International Euro-Par Conference, Aug 2004, Pisa/Italy, pp.272-283. ⟨inria-00001056⟩