基于GPU的串匹配算法研究  

String matching algorithm research based on graphic processing unit

作　　者：张庆丹[1] 戴正华[1] 冯圣中[1] 孙凝晖[1] 

机构地区：[1]中国科学院计算技术研究所

出　　处：《计算机应用》2006年第7期1735-1737,共3页journal of Computer Applications

基　　金：国家自然科学基金资助项目(60372040);中国科学院知识创新工程重大项目(KSCX2-SW-223)

摘　　要：BF算法是串匹配算法中最基础的算法,但它是串行算法,不适合图形处理器(GraphicProcessing Unit,GPU)的体系结构。结合GPU的特殊体系结构,通过数据存取方式和计算策略的改进,充分利用了GPU的并行处理能力,从而基于GPU实现了BF算法。实验结果表明基于GPU的并行算法能够取得较好的加速比,同时也给出了在现有GPU架构上有效实现通用计算的瓶颈。BF algorithm is the basic algorithm in string matching field, but it is not parallel, and does not match the system architecture of the Graphic Processing Unit（ GPU）. Concerning the specific hardware architecture of the GPU, new method of data accessing and computing strategy of the parallel BF algorithm was designed. And by fully using the parallel capaticity of the GPU, the GPU-based algorithm can outperform the CPU-based algorithm. The bottlenecks of GPU architectures were also presented.

关 键 词：图形处理器 通用计算 串匹配 并行 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]