Publication detail

Parallel Genetic Algorithm on the CUDA Architecture

POSPÍCHAL, P. JAROŠ, J. SCHWARZ, J.

Original Title

Parallel Genetic Algorithm on the CUDA Architecture

English Title

Parallel Genetic Algorithm on the CUDA Architecture

Type

conference paper

Language

en

Original Abstract

This paper deals with the mapping of the parallel island-based genetic algorithm with unidirectional ring migrations to nVidia CUDA software model. The proposed mapping is tested using Rosenbrock's, Griewank's and Michalewicz's benchmark functions. The obtained results indicate that our approach leads to speedups up to seven thousand times higher compared to one CPU thread while maintaining a reasonable results quality. This clearly shows that GPUs have a potential for acceleration of GAs and allow to solve much complex tasks.

English abstract

This paper deals with the mapping of the parallel island-based genetic algorithm with unidirectional ring migrations to nVidia CUDA software model. The proposed mapping is tested using Rosenbrock's, Griewank's and Michalewicz's benchmark functions. The obtained results indicate that our approach leads to speedups up to seven thousand times higher compared to one CPU thread while maintaining a reasonable results quality. This clearly shows that GPUs have a potential for acceleration of GAs and allow to solve much complex tasks.

Keywords

massively parallel, genetic algorithm, island model, CUDA, migrations

RIV year

2010

Released

09.04.2010

Publisher

Springer Verlag

Location

Berlin Heidelberg

ISBN

978-3-642-12238-5

Book

Applications of Evolutionary Computation

Edition

Lecture Notes in Computer Science

Edition number

NEUVEDEN

Pages from

442

Pages to

451

Pages count

10

URL

BibTex


@inproceedings{BUT34649,
  author="Petr {Pospíchal} and Jiří {Jaroš} and Josef {Schwarz}",
  title="Parallel Genetic Algorithm on the CUDA Architecture",
  annote="This paper deals with the mapping of the parallel island-based genetic algorithm
with unidirectional ring migrations to nVidia CUDA software model. The proposed
mapping is tested using Rosenbrock's, Griewank's and Michalewicz's benchmark
functions. The obtained results indicate that our approach leads to speedups up
to seven thousand times higher compared to one CPU thread while maintaining
a reasonable results quality. This clearly shows that GPUs have a potential for
acceleration of GAs and allow to solve much complex tasks.",
  address="Springer Verlag",
  booktitle="Applications of Evolutionary Computation",
  chapter="34649",
  edition="Lecture Notes in Computer Science",
  howpublished="online",
  institution="Springer Verlag",
  year="2010",
  month="april",
  pages="442--451",
  publisher="Springer Verlag",
  type="conference paper"
}