Study On Detection Technology Of Program Exception Based On DSP

At precent, application program in DSP-based embedded system mainly uses C/C++ language. Due to the characteristics of C language and an embedded operating system, lots of exceptions in the program can't be found by compiler, while they will appear during program running. These exceptions often result in system instability even system failures. Therefore, the technologies of program exception detection are very necessary.This paper mainly discusses about code rewrite, array and stack overflow. A detection system of DSP program exception is estabilshed using hardware and software methods. Thus it can find out many exceptions in progam.First, exception detection in DSP internal memory is realized using memory protection mechanism, and the sysetem has been widely used in practice. Actual application has proved that it's sensitive to array and pointer overflow, and it has successfully discovered many of these issues.Secondly, advanced event trigger is used in real system by software programming based on which exception detection in DSP external memory and program tracing functions are realized, so that it can detect exceptions at anytime. The former function can find out lots of exception accesses to external memory, while the data recorded by latter function have great effect on the analysis of exception.Finally, BP neural network algorithm is realized on DSP for stack exception detection of multi-task system. The BP algorithm has good function on complex classification problem. The algorithm can distinguish whether the stack exception occurs in multi-task system by learning stack samples.All detection technologies are verified by actual cases, and some of them have been widely used in real system. They resolve lots of potential problems and the stability of program has been greatly enhanced.