Communication efficient work distributions in stencil operation based applications

摘要

In recent years, the use of accelerators in conjunction with CPUs, known as heterogeneous computing, hasrnbrought about significant performance increases for scientific applications. One of the best examples ofrnthis is lattice quantum chromodynamics (QCD), a stencil operation based simulation. These simulationsrnhave a large memory footprint necessitating the use of many graphics processing units (GPUs) in parallel.rnThis requires the use of a heterogeneous cluster with one or more GPUs per node. In order to obtainrnoptimal performance, it is necessary to determine an efficient communication pattern between GPUs onrnthe same node and between nodes. In this paper, we present a performance model based method for minimizingrnthe communication time of applications with stencil operations, such as lattice QCD, on heterogeneousrncomputing systems with a non-blocking InfiniBand interconnection network. The proposedrnmethod is able to increase the performance of the most computationally intensive kernel of lattice QCDrnby 25% due to improved overlapping of communication and computation. We also demonstrate that thernaforementioned performance model and efficient communication patterns can be used to determine a costrnefficient heterogeneous system design for stencil operation based applications.