Workload decomposition strategies for hierarchical distributed‐shared memory parallel systems and their implementation with integration of high‐level parallel languages

Abstract

SUMMARYIn this paper we address the issue of workload decomposition in programming hierarchical distributedshared memory parallel systems. The workload decomposition we have devised consists of a two-stage procedure: a higher-level decomposition among the computational nodes; and a lower-level one among the processors of each computational node. By focusing on porting of a case study particle-in-cell application, we have implemented the described work decomposition without large programming effort by using and integrating the high-level language extensions High-Performance Fortran and OpenMP.

Hierarchical MPI+OpenMP Implementation of Parallel PIC Applications on Clusters of Symmetric MultiProcessors

“…In particular, we have designed and implemented the hierarchically combined particle-particle and particle-domain decomposition strategies, with the integrated use of HPF and OpenMP [2].…”

OhHelp

“…Thus a simple combination of static or dynamic domain decomposition and static particle decomposition, such as those exploiting [25] 2P/N 2D/N [21] P/N D OhHelp P/N 2D/N shared and distributed memory hierarchy [3,4,8] or dimension oriented ones [31], is not scalable because it cannot keep…”