Relationship analysis of three parallel codes

Abstract

The increasing demand for computing capability is driving the appearance of diverse parallel paradigms. Recently, MPI+OpenMP hybrid paradigm codes have been proved high performance efficiency in many applications and have been widely studied. But its relationship with MPI and OpenMP paradigm codes still need further study. In this paper, we focus on two aspects, the overhead and the number of processors definition methodology, to discuss the relationship between MPI, OpenMP and MPI+OpenMP paradigm codes. This paper provides a collection of experimental results of these three paradigm codes. The experimental results demonstrate that the parallel efficiency of MPI+OpenMP hybrid paradigm code can be more properly figured out but its performance would not always be the best on SMPs, which means the application situation should be pre-studied.