Back to Search Start Over

Parallelizing Applications with Data-dependency with AOMP in a P2P-based High Performance Computing Platform.

Authors :
Luo, Feil
Li, Xiaoyong
Chen, Lin
Source :
Energy Procedia; Dec2011, Vol. 13, p7847-7855, 9p
Publication Year :
2011

Abstract

Abstract: An asynchronous one-sided message-passing programming (AOMP) model is provided in a in a P2P-based high performance computing system, P2HP-2. Here the instructions to parallelize applications with AOMP in P2HP-2 are presented in this paper. Therefore, the characteristics of applications are first formulized, where those with static datadependency can currently be applied in P2HP-2. Then the application are programmed with AOMP as a project with simple program multiple data (SPMD), where the procedure is processed as pretreatment, distributed computing, and post-treatment. Utilizing the application programming interfaces (APIs) of AOMP, the project is programmed and can be submitted to P2HP-2. Finally, a parallel serial threading algorithm with static data-dependency is created with AOMP. Through the experiments and analysis, great parallel efficiency has been achieved with the AOMP in P2HP-2, especially for those applications with static data-dependency [Copyright &y& Elsevier]

Details

Language :
English
ISSN :
18766102
Volume :
13
Database :
Supplemental Index
Journal :
Energy Procedia
Publication Type :
Academic Journal
Accession number :
85749555
Full Text :
https://doi.org/10.1016/j.egypro.2011.12.529