基于GPU的MATLAB计算与仿真研究  

Research of MATLAB Calculations and Simulation based on GPU

作　　者：王恒[1] 高建瓴[1] 

机构地区：[1]贵州大学计算机科学与信息学院,贵州贵阳550025

出　　处：《贵州大学学报（自然科学版）》2012年第6期95-98,112,共5页Journal of Guizhou University:Natural Sciences

基　　金：贵州省教育厅一般项目(黔教科(2010006)号)

摘　　要：图形处理单元(GPU)已经成为当今的主流计算系统的一个组成部分,现代GPU不仅是一个功能强大的图形引擎,也是一个高度并行的可编程处理器,GPU的峰值运算和内存带宽往往大幅超出其CPU所对应的峰值和内存带宽。本文介绍了基于GPU通用计算框架的JACKET加速MATLAB的计算仿真方法,通过FFT算法得出仿真结果,分析在CPU和GPU运行环境下的GFLOPS和加速比,最后得出基于GPU的MATLAB计算仿真程序运行效率在JACKET的加速下大大提高了。The graphics processing unit （GPU） has become tern, modern GPU is not only a powerful graphics engine, an integral part of today mainstream computing sys- but also a highly parallel programmable processor. Peak arithmetic and memory bandwidth of the GPU often significantly exceeded the corresponding peak and mem- ory bandwidth in its CPU. This paper describes a computational simulation method based on GPU general compu- ting framework for the JACKET accelerating MATLAB, draws simulation results through classic FFT algorithm. analyzes the FFT algorithm in the CPU and GPU running environment Gflops and speedup and comes to the conclusion that under JACKET large scale computing and the GPU-based MATLAB simulation program running has been efficiency greatlv imoroved.

关 键 词：GPU(Graphics Processing Unit) CUDA(Compute UNIFIED DEVICE Architecture) JACKET 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]