CPU And GPU Mixed Architecture Of Underwater Array Processing Software Design And Implementation

With the ocean exploration and development deepening in contemporary society,ocean battlefield is receiving more and more attention as an important strategic development object.At the same time as the acoustic technology and electronic information technology is developing rapidly,underwater weapons have been replaced by new ones.Sonar plays a crucial role in naval warfare.Various Marine environment makes sonar’s work environment increasingly complex,the signal processing which sonar internal involves is complicated.The underwater array detection technology is the core implementation of sonar,so achieving the real-time processing of the underwater array detection technology is increasingly prominent,speeding signal processing of underwater array detection system is urgently needed to solve.In the development of electronic information technology in recent years,GPU parallel computing is getting promoted and CUDA parallel architecture has also been updated.Parallel processing is widely used in the field of radar signal real-time processing that provides the research thought for the underwater array signal processing,through the study found that there has a lot algorithms in the underwater array signal processing are suit for parallel computing.Using the GPU parallel computing to speed up,thus we can improve the efficiency of the whole underwater array detection system.Taking the underwater array detection system as the research object,firstly analyzing the underwater array detection system’s developmental requirements,the system is divided into three software programs,which are signal transmitting software,array signal processing software and displaying software,between different softwares through the network to transmitting data,from the system level proposing the overall implementation of underwater array detection system basing on CPU and GPU collaborative architecture.Combining multiple array signal processing beamforming algorithms,firstly,making an introduction of the principle of array signal beamforming algorithms,analyzing the operation characteristics of each algorithm.Then,analyzing the storage model of CUDA platform as well as its own library functions and so on,completing the design of parallel array signal beamforming algorithms and programming logic on GPU platform.Finally,on the CPU and GPU platform,simulating algorithms by the methods of control variables to test and compare algorithm’s performance.At last integrating all the software modules as a complete underwater array detection system,using Labwindow/CVI to realize the terminal display software,and using the experiment data to test system’s performance,which verifies function of the real-time processing of underwater array detection system.This article uses the CPU and GPU mixed architecture to build a model of underwater array detection system,the making the distribution that the CPU is responsible for the system logic control and task assignment,GPU is in charge of large-scale data parallel computation,implementing the tasks parallel.Using multi-threaded parallel tasks,asynchronous concurrent streaming and hardware devices in parallel mode to implement multiple modules parallel,by calling the CUDA to make algorithms’ internal parallel multi-thread.From the levels of tasks,modules,and algorithms realizing the real-time processing of the system.Among them,the array signal processing software can carry on the accurate and quick calculation.Display software can show data in real time,at the same time takes abilities for data processing,and has the good man-machine interacting effection,which has certain reference significance for the development of underwater array detection technology.