Program Real-Time Performance Analysis Based On Stateflow Models

The enhancement of complexity in embedded systems pushes more system designers to do system planning with model-based development environments.As one of the most widely applied development environments based on models,the combination of MATLAB/Simulink/Stateflow can be found in various fields such as embedded system design.Obviously,real-time performance analysis of embedded systems modeled by Stateflow is of great practical meaning.Stateflow is based on the model of finite-state machine,so it has strong expressiveness of systems.As a result,it is almost impossible to analyze its schedulability precisely in most situations.A feasible method is to modify existed models with weaker expressiveness and lower analysis complexity to approach Stateflow model.Our research target is to analyze real-time performance of Stateflow with state transition conditions according to current DRT task model.To obtain the goal,we firstly do research about abstraction interpretation theory and interval power set abstraction domain theory,and propose the computation formula of simple logical expressions of interval power set.According to theories above,we modify the existed value range analysis algorithm and put forward our new algorithm for C program,which can not only improve the precision of the previous one but make analysis process simulate program execution sequence,so that we can get intuitive analysis process and results.After that,we design and realize a tool named interVALue to do value range analysis for C program.Now armed with the ability to analyze the value range of C program automatically,we get a deeper understanding of DRT model and improve it as well as real-time performance analysis methods in accordance with target Stateflow model.At the same time,we propose DRT-G model,analyzing its description ability and release sequence structure.According to the characteristics of the model,we present how to do real-time performance analysis with DRT-G model,and improve the analysis efficiency of related methods.After modifying analysis tool of DRT model according to DRT-G model,we compare and analyze the results generated by both new and previous methods for a system instance.Results show that our DRT-G model as well as its analysis methods performs well when modeling and analyzing the real-time performance of Stateflow model with state transitions,producing a precise judgment of system schedulability.However,DRT model lacks enough expressiveness since it does not describe state transition conditions.When applied to the system instance above,DRT task model may generate imprecise even wrong results.