A loop unrolling method based on machine learning

In order to improve the accuracy of loop unrolling factor in the compiler, we propose a loop unrolling method based on improved random decision forest. First, we improve the traditional random decision forest through adding weight value. Second, BSC algorithm based on SMOTE algorithm is proposed to...

Full description

Saved in:
Bibliographic Details
Published inVibroengineering Procedia Vol. 18; pp. 215 - 221
Main Authors Liu, Hui, Guo, Zhanjie
Format Journal Article
LanguageEnglish
Published JVE International Ltd 01.05.2018
Subjects
Compilers (Software)
Machine learning
Methods
Optimization theory
Testing
Online AccessGet full text

Cover

More Information
Summary:In order to improve the accuracy of loop unrolling factor in the compiler, we propose a loop unrolling method based on improved random decision forest. First, we improve the traditional random decision forest through adding weight value. Second, BSC algorithm based on SMOTE algorithm is proposed to solve the problem of unbalanced data sets. Nearly 1000 loops are selected from several benchmarks, and features extracted from these loops constitute the training set of the loop unrolling factor prediction model. The model has a prediction accuracy of 81 % for the unrolling factor, and the existing Open64 compiler gives 36 % only.
ISSN:2345-0533
2538-8479
DOI:10.21595/vp.2018.19928