Performance Evaluation Of Algorithm For Recognition Of Targets In The Sky On Leon3 Multi-core SoC

With the rapid development of science and technology,SoC has been widely used in automobile manufacturing,industrial control,intelligent equipment and other fields,becoming the mainstream of today's technological development.It is an integrated circuit chip,which can reduce system development costs and shorten the development cycle.However,with the increasing amount of information,single-core CPU has been difficult to meet the demand in many high demand occasions.people hope that embedded devices have a higher performance.As a result,the demand for multi-core SoC is becoming stronger.This paper first introduces the current situation of SoC technology development at home and abroad,lists some representative products,and analyzes its function and application.This paper also compares the popular soft-core CPU architecture,and selects LEON series 32-bit CPU developed by Gaisler Reserch as the soft-core CPU of the system.It supports the GRLIB IP library,with a perfect development environment,which greatly facilitates the design of SoC;Secondly,this article also compares today's popular open source embedded real-time operating system,and has a research on the popular eCos operating system,which is an open source software,characterized by completely free,with no copyright restrictions;Thirdly,this paper introduces the flow of the algorithm for target recognition,analyzes the principle of the algorithm and the possibility of parallelization,and designs the parallelization of the part which can be parallelized according to the multi-thread and synchronization mechanism of eCos;Finally,single-core,dual-core and quad-core SoC based on the leon3 architecture are built on the FPGA development board,the speed of the algorithm for recognition of targets in the sky execution was tested and compared and the factors that affect the performance of parallelization was analyzed.The results show that multi-core SoC has obvious advantages in computing speed compared with single-core system.The degree of parallelization of the algorithm and the size of the cache will have an impact on the effect of parallelization,thus affecting the speed of software execution.