基于GPGPU的生物序列快速比对

摘要

This paper presents a highly effective biological sequences alignment solution based on CPU-Graphic Processingr Unit(GPU) heterogeneous platform. The solution re-write Smith-Waterman(SW) algorithm in OpenCL platform, reorganizes the algorithm to adapt the GPU architecture and applies optimization of memory accessing latency and optimization of the memory deployment, fully utilizes GPU's parallel processing capability, greatly improves the effectiveness of algorithm. Experimental result shows that the solution has 100 times performance increase than the serial algorithm in CPU.%在CPU-GPU异构平台下,提出一种高效的生物序列比对方案.该方案利用GPU的并行处理能力,通过对读延迟、写延迟、重组函数及数据传输进行优化,在OpenCL框架下重构Smith-Waterman算法,加快生物序列比对速度.实验结果证明,与CPU上传统的串行算法相比,该算法最高可获得约100倍的性能提升.