Porting And Optimizing SOAP2 On Loongson Platform

Biological sequence alignment tools is used to explore biological sequences(DNA sequence、RNA sequence and protein sequence etc.) functional, structural and evolutionary information. However, a lot of alignment tools cannot meet the growing requirement of matching speed, accuracy and sensitivity etc., when encountered massive sequence data generated by a new generation of sequencing technology.There are two main kinds of algorithms for the implementation of alignment tools. One is based on hash index and combined with dynamic programming algorithm. The other one is based on FM-index and combined with BWT matching method. The former speed is lower compared with the latter, but has a higher precision, sensitivity and more reported matched hits etc., The latter speed is higher compared with the former, but has a lower precision, sensitivity and reported match hits.This paper focuses on software optimization to study biological sequence alignment technology, and take mainstream sequence alignment tool SOAP2 as optimization object on Loongson platform. Firstly, the research background and current research situation of biological sequence alignment tools are introduced. Then this paper analyzes the necessity of porting and optimizing SOAP2 on Loongson platform. Finally, three types of transplantation and optimization problems of SOAP2 are analyzed and their own corresponding solutions are proposed.The contributions and work of this paper are as follows:1. For the problem of that SOAP2 cannot run on Loongson platform directly, the SOAP2 is transplanted and optimized comprehensively for the first time. This paper also put forwards a method for more efficient use of Loongson multimedia instructions and replaces the SSE3 instructions in the SOAP2 with Loongson multimedia instructions by this method, not only achieving the transplantation, but also improving its performance.2. For the problem of that SOAP2’s speed caused by the bottlenecks of IO, the thread pool is used to optimize the software IO to increase matching speed. Thread pool is mainly to solve the problem of the long-waiting time of the child threads waiting main thread reading and writing IO, and in order to acquire a higher performance in multi-core machine.3. For the problem of sensitivity, reported matched hits and multi mapped rate, this paper put forwards a new algorithm based on BWT search algorithm and combination idea and the SOAP2 is optimized by this algorithm. The algorithm can better take advantage of hash index and FM-index. It can maintain the similar matching speed, but enhance the sensitivity, reported matched hits and multi mapped rate, and also reduces the memory overhead.