A study on popular auto‐parallelization frameworks

Summary We study five popular auto‐parallelization frameworks (Cetus, Par4all, Rose, ICC, and Pluto) and compare them qualitatively as well as quantitatively. All the frameworks primarily deal with loop parallelization but differ in the techniques used to identify parallelization opportunities. Due...

Full description

Saved in:
Bibliographic Details
Published inConcurrency and computation Vol. 31; no. 17
Main Authors Prema, S., Nasre, Rupesh, Jehadeesan, R., Panigrahi, B.K.
Format Journal Article
LanguageEnglish
Published Hoboken Wiley Subscription Services, Inc 10.09.2019
Subjects
Dependence
loop transformations
loop‐carried dependence
privatization
vectorization
Online AccessGet full text

Cover

More Information
Summary:Summary We study five popular auto‐parallelization frameworks (Cetus, Par4all, Rose, ICC, and Pluto) and compare them qualitatively as well as quantitatively. All the frameworks primarily deal with loop parallelization but differ in the techniques used to identify parallelization opportunities. Due to this variance, various aspects, such as certain loop transformations, are supported only in a few frameworks. The frameworks exhibit varying abilities in handling loop‐carried dependence and, therefore, achieve different amounts of speedup on widely used PolyBench and NAS parallel benchmarks. In particular, Intel C Compiler (ICC) fares as an overall good parallelizer. Our study also highlights the need for more sophisticated analyses, user‐driven parallelization, and meta‐auto‐parallelizer that provides combined benefits of various frameworks.
ISSN:1532-0626
1532-0634
DOI:10.1002/cpe.5168