| | |
Summary: Predicting Unroll Factors Using Supervised Classification
Mark Stephenson and Saman Amarasinghe
Massachusetts Institute of Technology
Computer Science and Artificial Intelligence Laboratory
Cambridge, Massachusetts
{mstephen,saman}@mit.edu
Abstract
Compilers base many critical decisions on abstracted ar-
chitectural models. While recent research has shown that
modeling is effective for some compiler problems, building
accurate models requires a great deal of human time and ef-
fort. This paper describes how machine learning techniques
can be leveraged to help compiler writers model complex
systems. Because learning techniques can effectively make
sense of high dimensional spaces, they can be a valuable
tool for clarifying and discerning complex decision bound-
aries. In this work we focus on loop unrolling, a well-known
optimization for exposing instruction level parallelism. Us-
ing the Open Research Compiler as a testbed, we demon-
strate how one can use supervised learning techniques to
|
