부산대학교 도서관

In the design of trustworthy software for real-time embedded systems, the interrupt mechanism plays an important role. Due to the randomness and non-determinism of interrupt handling behaviors, the performance evaluation of embedded software is an important but challenging problem. To solve this problem, we propose a performance model for embedded software based on extended deterministic and stochastic Petri nets (EDSPN), which is a semantic expansion of DSPN that add non-deterministic time transitions. In this manner, the performance model can effectively express interrupt handling behaviors, including random requests, responses according to priority, interrupt nesting, and non-deterministic execution times. In addition, we also provide a performance evaluation method for the EDSPN model based on Markov regenerative theory and demonstrate the analysis process via a specific example. Finally, we analyze the performance of embedded software, taking into account all types of the impact factors of interrupts through experimentation.