Characterizing the Effectiveness of Compilers in Vectorizing Polyhedrally Transformed Code

Many of the compute intensive applications spend most of their time inside nested loops. Hence optimization of these nested loops can provide significant improvements in the speed of the program. A number of optimizations can be performed to a program in order to speed it up on a particular hardware. Optimization techniques such as Tiling, Vectorization, Loop Unrolling etc. can produce significantly better performance. In this study we focus on Tiling and Vectorization. Our study is to evaluate for various benchmarks and various problem sizes, whether one optimization affects the other or not. In cases where one optimization negatively affects the other, we evaluate the extent of the negative effect, which gives us an understanding of the type of optimization that should be performed in order to get an overall gain in the speed. This study evaluates two tiling schemes namely, PLuTo and PTile, with two compilers namely, the GNU C Compiler and the Intel C Compiler.

Name thesis:degree_name

Master of Science

Level thesis:degree_level

masters

Discipline thesis:degree_discipline

Computer Science and Engineering

Grantor dc:publisher

The Ohio State University

Year dc:date

2013

Author dc:creator

• Chidambarnathan, Yogesh

Contributors dc:contributor

• Sadayappan, Ponnuswamy

• Computer Science
• Vectorization
• Tiling
• Polyhedral model
• PAPI
• Pluto
• Ptile

Statement dc:rights

• unrestricted
• This thesis or dissertation is protected by copyright: all rights reserved. It may not be copied or redistributed beyond the terms of applicable copyright laws.

Language dc:language

English

Repository record dc:identifier

http://rave.ohiolink.edu/etdc/view?acc_num=osu1357321939

OAI identifier oai:identifier

oai:etd.ohiolink.edu:osu1357321939