Research Of Control Flow Detection Algorithm Based On Highly Reliable R80515

Microprocessor is increasingly used in the harsh environment. Single Event Effects refer to the phenomena that a single high-energy particle interferes electronic devices, which is the major factor decreasing the reliability of microprocessor in the space environment. Control flow detection technology is one of the effective means to prevent the program running errors caused by Single Event Upset.At present, there are two ways of control-flow detection: software way and hardware-software way. The software way is flexible and suitable for different hardware platforms, but it has a large system overhead. Contrastively, the hardware-software way has a small system overhead and a broad application prospect. This dissertation does research on the highly reliable R80515 which is designed by National Defense Science and Technology University.The dissertation firstly introduces the control-flow detection theory and several common control-flow detection algorithms, emphasizes the assembly-level control-flow detection method-CFCSS, and explicates control-flow detection principle, signature value assigned algorithm, detection process of control-flow and other technical details. Then the confusion problem and multiple sharing-in nodes in control-flow detection algorithm is pointed out, and the wrong type of control-flow can not be detected by CFCSS algorithm are classified. Furthermore, the dissertation proposes CFCMAS control-flow detection algorithm based on multi-adjustment signature by improving the basic block partition method, the signature value generation method and the signature value assigned algorithm. The new control-flow detection algorithm solves the inherent drawbacks existed in CFCSS algorithm and saves the storage space and time consumption greatly.On the basis of fundamental research, the dissertation implements CFCMAS algorithm by inserting a control-flow detection layer in the general embedded compiler SDCC, chooses several programs for testing according to different control-flow detection algorithm, and differentiates CFCMAS algorithm from the other common control-flow detection algorithms on the performances in the experiment. The experiment results show that the performance and resource consumption of CFCMAS algorithm are superior to other algorithms.