| With improvement of the scale and complexity of the embedded software. the quality of embedded software play a decisive role in the quality of the embedded system.and the performance of embedded software directly influence the quality of applications,so all the embedded software developer want to need one performance testing tool, which can test the performance of embedded software in specific domain.The tool can analyse the performance of the embedded software,and the developers can easily find the weaknesses which affect the embedded software quality,and then the developers modify the codes to improve the quality of software.In this dissertation, we analyse the main characters of some tradition embedded software performance testing tool, and we study the basic methods used to test embedded software performance.The performance testing tool described in this paper is designed for the computation-intensive embedded software,and dynamicily analyse the function performance characteristics of software in the execution.We design the performance tool which is based on gprof, relative to the gprof,the tool can test the dynamic function and static function which include multi-threads and multi-task the software.We design the software architecture and internal modules of the performance testing tool, and study all the modules and interrelation among each module. Relative to gprof.We design and implement the new arithmetic of performance data collection module and data communication module,and improve the performance data analysis module of gprof.First of all,study the performance data collection module.Different performance data represents the different function type,and we are used to insert stub-code of the source codes,installing breakpoint technique and so on, and collect the performance data.We study the performance data storage form,and design and implement the algorithm which can quickly find and keep performance data.We study multi-processes programming method,and use the process trace technique to implement the multi-processes performance data collection.Secondly,in the way of analyzing the performance data,we mainly study the the algorithm which is used to transform the performance data based on the analysis of gprof performance data.Through modifying the old algorithm,we design and implement the performance data analysis which provides for the dynamic function and multi-process testing.At last,we study the cross-development character of the performance testing of embedded software,and we design and implement the target communication module and the host communication module. Designing the simple and reliable coding strategy, which transfers the different types of performance data.We use the OS API to set the communication configuration and implement the controlled and reliable transmit the data between the host and the target.Depending on the environment and testing objective of performance testing tool,we design the testing case.In the face of the performance testing tool,We implement the testing case and analyse the result, to test the basic function of the performance testing tool.Compared to the other performance testing tools,we demonstrate the tool usable and practical. |