An extended OpenMP targeting on the hybrid architecture of SMP-cluster

Authors:
Zhao Yudi;Hu Changjun;Wang Shengyuan;Zhang Suqin
Affiliations:
Dept. of Computer Science and Technology, Tsinghua University, Beijing, China;Information Engineering School, University of Science and Technology Beijing, Beijing, China;Dept. of Computer Science and Technology, Tsinghua University, Beijing, China;Dept. of Computer Science and Technology, Tsinghua University, Beijing, China
Venue:
ACST'06 Proceedings of the 2nd IASTED international conference on Advances in computer science and technology
Year:
2006

Citing 6
Cited 0

Optimizing for parallelism and data locality

ICS '92 Proceedings of the 6th international conference on Supercomputing
Language and Compiler Support for Hybrid-Parallel Programming on SMP Clusters

ISHPC '02 Proceedings of the 4th International Symposium on High Performance Computing
Communication Bandwidth of Parallel Programming Models on Hybrid Architectures

ISHPC '02 Proceedings of the 4th International Symposium on High Performance Computing
Program Development Environment for OpenMP Programs on ccNUMA Architectures

LSSC '01 Proceedings of the Third International Conference on Large-Scale Scientific Computing-Revised Papers
Compiler Optimization of Implicit Reductions for Distributed Memory Multiprocessors

IPPS '98 Proceedings of the 12th. International Parallel Processing Symposium on International Parallel Processing Symposium
Improving the performance of OpenMP by array privatization

WOMPAT'03 Proceedings of the OpenMP applications and tools 2003 international conference on OpenMP shared memory parallel programming

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper, we present an Extended OpenMP language (EOMP) and its hybrid execution model on SMP cluster. In order to provide programmers with a simple, flexible and relatively efficient parallel paradigm for SMP cluster, we extend OpenMP with new directives for computing resource definition and data mapping. EOMP supports both task parallelism and data parallelism on SMP cluster, while enable the programmers to exploit both internode and intra-node data locality. A runtime library for EOMP has been developed. Experiments based on this library prove the feasibility of the execution model, and show performance speedups gained from exploiting intra-node data locality.