Abstract:
This paper describes the acceleration of hmmsearch by exploiting fine-grained designs based on FPGA. It exploits data reuse schemes by making use of the method of fetching, slip windows and data transmition and utilizes pipeline in the matrix cell. Experimental results show the performance can reach to 3.6 GCUPS, which is nearly 235 speedup over that of CPU. One PE calculating performance improves 34% over the fastest FPGA accelerator of the same kind.
Key words:
biological sequences search,
accelerator,
FPGA
摘要: 在FPGA平台实现细粒度并行的hmmsearch加速技术。采用数据预取、滑动窗口和数据传递等策略实现子处理单元的数据重用。在计算矩阵块内部实现流水线计算。加速器性能为3.59 GCUPS,与CPU相比,可获得接近235倍的加速效果。与目前FPGA上同性质最快的加速器相比,单PE可获得34%的性能提升。
关键词:
生物序列搜索,
加速器,
现场可编程门阵列
CLC Number:
LI Rong-Chun, DOU Yong, JIA Fei. Acceleration Technology of Biological Sequences Search Algorithm hmmsearch[J]. Computer Engineering, 2010, 36(20): 265-267.
李荣春, 窦勇, 夏飞. 生物序列搜索算法hmmsearch的加速技术[J]. 计算机工程, 2010, 36(20): 265-267.