Performance modeling of hybrid MPI/OpenMP scientific applications on large-scale multicore supercomputers

作者：

Highlights：

• We present a performance modeling framework based on memory bandwidth contention time and a parameterized communication model.

• We model and predict the performance of OpenMP, MPI and hybrid scientific applications with weak scaling on multicore supercomputers.

• We use STREAM memory benchmarks to provide initial performance analysis and model validation of MPI and OpenMP applications.

• We also use a weak-scaling hybrid large-scale scientific application: Gyrokinetic Toroidal Code in magnetic fusion to validate the performance model.

摘要

•We present a performance modeling framework based on memory bandwidth contention time and a parameterized communication model.•We model and predict the performance of OpenMP, MPI and hybrid scientific applications with weak scaling on multicore supercomputers.•We use STREAM memory benchmarks to provide initial performance analysis and model validation of MPI and OpenMP applications.•We also use a weak-scaling hybrid large-scale scientific application: Gyrokinetic Toroidal Code in magnetic fusion to validate the performance model.

论文关键词：Performance modeling,Hybrid MPI/OpenMP,Multicore supercomputers,Memory bandwidth contention time

论文评审过程：Received 20 December 2011, Revised 12 May 2012, Accepted 22 February 2013, Available online 13 March 2013.

论文官网地址：https://doi.org/10.1016/j.jcss.2013.02.005