基于定制协处理器的基因重测序加速技术研究

发布时间：2018-04-05 17:28

本文选题：高通量测序技术　切入点：短序列比对　出处：《计算机研究与发展》2014年09期

【摘要】：自2008年1月高通量测序技术应用以来,测序的通量和成本都在不断下降.然而基因数据的爆发式增长速度已经超过了摩尔定律,对海量数据的计算处理能力成为制约基因测序应用推广的瓶颈.以基于Hash索引的重测序算法为目标,对计算和访存行为进行分析,从而提出了一个现场可编程门阵列(field programmable gate array,FPGA)作为协处理器的架构,并在Convey公司的HC-1ex平台上进行了设计与实现.其基本处理单元内部采用全流水的设计及FIFO隔离计算模块和访存模块,可以完整执行重测序算法的核心流程.通过将基本处理单元和访存端口的一对一绑定,在4块Xilinx Virtex-6LX760上实现了64路并行处理流程,总平均读内存带宽可达22.59GBps.与8核Intel Xeon处理器相比,可以提升28.5倍的性能.
[Abstract]:Since the application of high throughput sequencing technology in January 2008, the throughput and cost of sequencing have been decreasing.However, the explosive growth rate of gene data has exceeded Moore's law, and the computing ability of mass data has become the bottleneck of gene sequencing application.Based on the Hash index based resequencing algorithm, this paper analyzes the behavior of computing and memory access, and proposes a field programmable gate FPGA (FPGA) as a coprocessor, which is designed and implemented on the HC-1ex platform of Convey.In the basic processing unit, the core flow of the resequencing algorithm can be carried out completely by using the design of all income, the FIFO isolation calculation module and the memory access module.By binding the basic processing unit to the memory access port one-to-one, a 64-channel parallel processing flow is implemented on 4 Xilinx Virtex-6LX760. The total average read memory bandwidth can reach 22.59 GB / s.Compared with 8-core Intel Xeon processor, it can improve 28.5 times performance.
【作者单位】：中国科学院计算技术研究所高性能计算机研究中心;中国科学院大学;
【基金】：国家“九七三”重点基础研究发展计划基金项目(2012CB316502)
【分类号】：Q75;TP332

【相似文献】