English

MOM: Matrix Operations in MLIR

Programming Languages 2022-08-23 v1 Mathematical Software

Abstract

Modern research in code generators for dense linear algebra computations has shown the ability to produce optimized code with a performance which compares and often exceeds the one of state-of-the-art implementations by domain experts. However, the underlying infrastructure is often developed in isolation making the interconnection of logically combinable systems complicated if not impossible. In this paper, we propose to leverage MLIR as a unifying compiler infrastructure for the optimization of dense linear algebra operations. We propose a new MLIR dialect for expressing linear algebraic computations including matrix properties to enable high-level algorithmic transformations. The integration of this new dialect in MLIR enables end-to-end compilation of matrix computations via conversion to existing lower-level dialects already provided by the framework.

Keywords

Cite

@article{arxiv.2208.10391,
  title  = {MOM: Matrix Operations in MLIR},
  author = {Lorenzo Chelini and Henrik Barthels and Paolo Bientinesi and Marcin Copik and Tobias Grosser and Daniele G. Spampinato},
  journal= {arXiv preprint arXiv:2208.10391},
  year   = {2022}
}

Comments

3 pages, 1 figure, 1 table, and 3 listings. Short paper presented at 12th International Workshop on Polyhedral Compilation Techniques (IMPACT 22)

R2 v1 2026-06-25T01:52:33.758Z