Research On SmithWaterman And HEVC Software Based On CPU-GPU Heterogeneous Parallel System

In recent years,with the development of GPU,the CPU-GPU heterogeneous system is gradually used by people.The CPU is responsible for the high complexity of the logical operation and GPU is responsible for the calculation of large data parallel work.The two systems cooperate with each other,and the system's overall computing power makes the most use.The applications based on CPU-GPU heterogeneous system come more,but the process of porting is complex and the result is not accurate.It`s still a long way before the CPU-GPU heterogeneous system could be used in porting large scale programs.Therefore,based on the original GPU acceleration research,this paper summarize the current two porting methods called MOL and RPA.And then this paper propose a new porting method called ISC.With this new method,the biological sequence alignment tools,SmithWaterman and BLAST,have been ported to GPU.Firstly,the paper introduces the hardware structure of GPU and the architecture of CUDA,and introduces the composition of CPU-GPU heterogeneous parallel system.This paper puts forward the general steps of the porting method,and summarizes two kinds of code porting methods,MOL and RPA.The two methods require programmers to have an in-depth understanding of the original CPU code,and because of the modification of the original code,it`s not only time consuming but also can`t guarante the result.In this paper,a new method of code porting ISC is proposed.The method does not need to make any changes to the original algorithm and data structure.It can save time and energy and can ensure that the results of the program are consistent with the results of the CPU version.And then,the ISC method is used to port the biological sequence alignment tool to its CUDA program and the result is given.The last part of this paper describes the latest video encoding standard HEVC and its CUDA version.The results show that with ISC method,CPU-GPU heterogeneous parallel systems can be effective for acclerating computing intensive software tools,greatly shorten the calculation time.