Title: HPM: a hierarchical model for parallel computations

Authors: Xiangzhen Qiao, Shuqing Chen, Laurence Tianruo Yang

Addresses: Institute of Computing Technology, Chinese Academy of Sciences, Beijing, 100080, PR China. ' Institute of Computing Technology, Chinese Academy of Sciences, Beijing, 100080, PR China. ' Department of Computer Science, St. Francis Xavier University, Antigonish, NS B2G 2W5, Canada.

Abstract: A hierarchical model for parallel computations is introduced and evaluated in this paper. This model describes the general homogeneous parallel computer systems with Hierarchical Parallelism and hierarchical Memories (named as HPM). The HPM model consists of a hierarchy of ERAMs that cooperate with each other. A parallel function HP describes the multi-level parallelism of the system. The memory function Hm shows the characteristics of the hierarchical memories. The organisation for implementing the memory accesses or data communication in a computer system is known by a joint name: generalised hierarchical memories. The binding relation HB gives the relation connecting parallelism and hierarchical memories. The HB defines the organisation of synchronisation between the subsystems, and forms the hierarchy relation tree. The performance of HPM algorithms is discussed. The |generalised locality| and |memory consistency| are proposed to analyse the algorithm performance. Their usage and examples are also given.

Keywords: HPM; ERAM; enhanced RAM; random access memory; binding relation; hierarchy relation tree; generalised locality; memory consistency; parallel computing; hierarchical parallelism; hierarchical memories; high performance computing; parallel processing.

DOI: 10.1504/IJHPCN.2004.007571

International Journal of High Performance Computing and Networking, 2004 Vol.1 No.1/2/3, pp.117 - 127

Published online: 05 Aug 2005 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article