APCM: An Auto-Parallelism Computational Model : Increasing the performance of MPI applications in multi-core environments

Given the availability of computer clusters based on multi-core processors, the hybrid programming model has become an important ally of high-performance computing users in improving the performance of their parallel applications. However, creating hybrid applications is a complex task because it re...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: Costa, André Luiz Lima da, Souza, Josemar Rodrigues de
Formato: Articulo
Lenguaje:Inglés
Publicado: 2013
Materias:
MPI
Acceso en línea:http://sedici.unlp.edu.ar/handle/10915/34513
http://journal.info.unlp.edu.ar/wp-content/uploads/JCST-Dec13-10.pdf
Aporte de:
Descripción
Sumario:Given the availability of computer clusters based on multi-core processors, the hybrid programming model has become an important ally of high-performance computing users in improving the performance of their parallel applications. However, creating hybrid applications is a complex task because it requires developers to be familiar with two distinct parallel programming models. Against this background, this article introduces APCM, an auto-parallelism computational model. APCM’s goal is to create hybrid parallel applications, i.e., OpenMP (memory programming) and a message-passing interface (MPI), from MPI applications. This goal is achieved in a simple, automated manner that is transparent for the user while increasing application performance. In the article’s conclusion, we present consistent results that attest the efficacy of the proposed model.