Loop Transformations: Convexity, Pruning and Optimization

Abstract : High-level loop transformations are a key instrument in mapping computational kernels to effectively exploit resources in modern processor architectures. However, determining appropriate compositions of loop transformations to achieve this remains a significantly challenging task; current compilers may achieve significantly lower performance than hand-optimized programs. To address this fundamental challenge, we first present a convex characterization of all distinct, semantics-preserving, multidimensional affine transformations. We then bring together algebraic, algorithmic, and performance analysis results to design a tractable optimization algorithm over this highly expressive space. The framework has been implemented and validated experimentally on a representative set of benchmarks run on state-of-the-art multi-core platforms.
Complete list of metadatas

Cited literature [49 references]  Display  Hide  Download

https://hal.inria.fr/inria-00551077
Contributor : Cédric Bastoul <>
Submitted on : Sunday, January 2, 2011 - 2:27:23 PM
Last modification on : Friday, April 27, 2018 - 2:42:02 PM
Long-term archiving on : Sunday, April 3, 2011 - 2:41:12 AM

File

PBBCRSV11-POPL.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : inria-00551077, version 1

Collections

Citation

Louis-Noël Pouchet, Uday Bondhugula, Cédric Bastoul, Albert Cohen, Jagannathan Ramanujam, et al.. Loop Transformations: Convexity, Pruning and Optimization. 38th ACM SIGACT-SIGPLAN Symposium on Principles of Programming Languages (POPL'11), Jan 2011, Austin, TX, United States. ⟨inria-00551077⟩

Share

Metrics

Record views

370

Files downloads

409