基于CMP+GPU的并行离散事件仿真技术初探

Prospect for Technologies of Parallel Discrete Event Simulation Based on CMP+GPU Heterogeneous Computing System

随着应用的不断深入,仿真对计算机综合性能的要求越来越高。异构计算系统的出现为并行离散事件仿真带来了重要的机遇和挑战,如何高效充分利用系统资源,提高仿真运行效率成为当前亟待解决的关键问题。从技术角度介绍了利用CMP和GPU进行并行仿真加速的研究进展,分析了传统高性能仿真技术难以充分利用异构计算系统资源的问题,在此基础上就如何利用异构计算系统进行并行离散事件仿真提出了具体建议和展望。

With the advance of applications, simulation has put heavier demand on computer's performance. The emergence of heterogeneous computing systems offers a new opportunity to satisfy such demands, but also brings some challenges. How to fully utilize the system resources to improve the efficiency of simulation becomes the key issue. The current researches were introduced about how to use CMP and GPU to accelerate parallel simulation, and the reasons why traditional methods of high performance simulation were not available were analyzed. Based on these analyses, some proposals and prospects were given on how to use heterogeneous computing system to accelerate parallel discrete event simulation.