Progress Report: A Deep Learning Guided Exploration of Affine Unimodular Loop Transformations

TL;DR

This paper introduces a deep learning-guided method for optimizing loop transformations in polyhedral compilers, achieving significant speedups by exploring affine and non-affine transformations.

Contribution

It presents a novel deep learning-based approach to automatically explore loop transformation sequences for improved program performance.

Findings

Achieves a 2.35x geometric mean speedup over Pluto

Guides transformation exploration with a deep learning cost model

Explores both affine and non-affine loop transformations

Abstract

In this paper, we present a work in progress about a deep learning based approach for automatic code optimization in polyhedral compilers. The proposed technique explores combinations of affine and non-affine loop transformations to find the sequence of transformations that minimizes the execution time of a given program. This exploration is guided by a deep learning based cost model that evaluates the speedup that each sequence of transformations would yield. Preliminary results show that the proposed techniques achieve a 2.35x geometric mean speedup over state of the art polyhedral compilers (Pluto).