BLAST加速资料.ppt

研究现状——NUDT 二 Seeds Merging: It is hard for ungapped extension stage to catch up with the speed of multi-seeds detect with the growth of the array size. The benefit of merging seeds is that the number of valid seed can be reduced significantly. 研究现状——NUDT 二 Multi-Seeds Extension: 研究现状——NUDT 二 研究现状——BSW 二 This paper presents an FPGA-based alignment tool that considerably accelerates the SW algorithm with a highly configurable and novel hardware computation engine, while providing a robust software interface that seamlessly integrates into existing BLAST pipelines. 研究现状——BSW 二 The novel hardware computation engine In this paper, we present the novel use of a method for incorporating control information into the data stream,limiting wasted cycles and increasing hardware utilization. 研究现状——BSW 二 Seq:{ACGT, TTG, ACNG, CNTG}. Set {1-2, 1-3, 1-4, 2-3, 2-4, and 3-4} The CPU send“LACGTRTTGPRACNGPRCNTGPLTTGRACNGPRCNTGPLACNGRCNTGPN” to the FPGA with a 3-bit encoding per character. After (50 + pipeline length) FPGA clock cycles, the six results are ready for retrieval when the CPU has free cycles. L：①配置每一个PE。②告知PE后面的序列只是加载。 R：①在比对之前Reset PE。②表明序列加载完成 P：①告知FIFO产生了结果。 研究现状——BSW 二 If the same loaded sequence is to be used more than once in successive comparisons, then it is not necessary to reload PEs after every comparison, and streamed sequences can be piped through the array one after another. LACGTRTTGPRACNGPRCNTGPLTTGRACNGPRCNTGPLACNGRCNTGPN where later PEs in the pipeline between an “L” and “R” can be reconfiguring while simultaneously an earlier group between an “R” and “P” can be processing a query while an even earlier group is doing something else. 研究现状——BSW 二 If the same loaded sequence is to be used more than once in successive comparisons, then it is not necessary to reload PEs after every comparison, and streamed sequences can be piped through the array one after another. LACGTRTTGPRACNGPRCNTGPLTTGRACNGPRCNTGPLACNGRCNTGPN where late