基于OpenCL的Smith-Waterman算法硬件加速研究  

Hardware acceleration of Smith-Waterman algorithm based on OpenCL

作　　者：申曦 蒋剑飞[1] SHEN Xi;JIANG Jian-fei(School of Microelectronics,Shanghai Jiaotong University,Shanghai 200240,China)

机构地区：[1]上海交通大学微电子学院,上海200240

出　　处：《信息技术》2020年第4期1-4,共4页Information Technology

基　　金：天津市自然科学基金(15JCQNJC00100)。

摘　　要：生物序列分析由于其数据的海量性、分析算法的多样性和复杂性,因此其对运算平台以及软件工具有着很高的要求。在生物序列分析领域中,文中针对序列比对所采用的经典算法即Smith-Waterman算法在FPGA加速平台下的性能进行研究,利用开放运算语言OpenCL进行异构平台的硬件加速设计。通过利用Smith-Waterman算法的波前特性,在硬件设计层面上实现算法在运算过程中的高度并行化,弥补了在CPU单一平台下只能进行串行运算的不足。通过对大量不同样本序列的测试表明,利用算法的波前特性,针对短序列比对,FPGA的运算速度最高能达到CPU的4倍。Due to the massive data and the diversity and complexity of the analysis algorithms,biological sequence analysis has high requirements on computing platforms and software tools.In the field of biological sequence analysis,the classical algorithm used in sequence alignment is used.The performance of Smith-Waterman algorithm under FPGA acceleration platform is studied,and the hardware acceleration design of heterogeneous platform is carried out by using open computing language OpenCL.By using the wavefront characteristics of Smith-Waterman algorithm,the algorithm is implemented in the hardware design level.The high degree of parallelism makes up for the insufficiency of serial operation in a single CPU platform.Tests on many different sample sequences show that,with the wavefront characteristics of the algorithm,for short sequence alignment,the FPGA can achieve up to four times the speed of the CPU.

关 键 词：异构加速 序列比对 史密斯沃特曼算法 波前特性 现场可编程逻辑阵列 

分 类 号：TN492[电子电信—微电子学与固体电子学]