Measuring the impact of Automatic Program Parallelization Techniques in Cetus v2.0
Date
2021-11-17
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Data Science Institute [DSI], University of Delaware, Newark, DE
Abstract
Cetus is a source-to-source translator for programs written in the C language. The primary use is as a parallelizing compiler, translating C programs to equivalent C code annotated with OpenMP parallel directives. Cetus is a research platform to study parallelization techniques and related program transformations. Cetus was created out of a need for a state-of-the-art automatic parallelizer for multicores, written in a modern language and capable of performing analyses and transformations for today’s architectures. This poster presents an in-depth evaluation of the existing and newly added analysis and transformation techniques in Cetus on a set of benchmark applications.