MPI-3 Shared Memory Programming Introduction

摘要: This chapter discusses an interprocess shared memory extension, which was added in MPI 3.0 standard, and how it can be used to enhance communication efficiency enable footprint optimization. A simple 1-D ring “hello world” example is shown then extended for several node runs. Using a modified MPPTEST benchmark, up 4.9× improvement over standard point-to-point approach on one Intel Xeon Phi coprocessor shown. have potential benefits with halo exchanges even multinode cases. Improvements 1.8× two coprocessors are along demonstrating that the of SHM grow increases number intranode neighbors exchanges.