Research On Periodic Scheduling Algorithm Verification Technology For RTEMS Application

RTEMS(Real Time Executive for Multiprocessor Systems) has always beenapplied in controlling system. It’s not only need the system could deal properly withexternal information but also focus on if the system could finish the task in certaintime with result, which different from other normal operating system. The externalinformation which need to be dealt may have no relation with each other, orsynchronized or exclusive with each other. The final aim of dealing with internalsystem task is correction and timely, if some of tasks can’t be finished in time mayeffects the whole executing process or even worse. Therefore, executing thedispatching test is necessary. The result of dispatching is closely relating todispatching algorithm. In order to guarantee the quality of real-time application, taskdispatching test is needed to prove real-time algorithm in the application.Usually speaking, cycling tasks is more common in controlling systems. Thereal-time dispatching algorithm has two mainly types, one is RMS another one isEDF. So, the following steps is my thesis mainly discuss on the basis of the twodispatching algorithm:(1) Under the circumstances of the loss caused by non-modifying applicationand minimum acquisition tasks, as well as by researching thread state extractiontechnology. Using GDB debugger especially for expanding RTEMS collect worsensituation in the operating process.(2) According task dispatching test and scheduling algorithm validation’s need,design strategy focus on applying RTEMS task status data, so as to minimize theperformance loss happened in the acquisition process and try to close to the crucialmoment of the task. Under the direction of data collecting strategy, the final result isthat the collecting rate could follow the task time of crucial to achieveauto-adjustment, use the optimal allocation and linear fitting problem to designingtask status acquisition algorithm.(3) Through the study on the model of finite-state machine, extend theautomatically machine of the finite state to design and achieve the time automaticmachine and Automatic Incident machine, which is to judge the state of tasks. Eventautomata machine may be changed by the result of collection, and then the Timeautomata manchine will changed by the event automata machine.Comparing the data collecting algorithm and equivalent periodic data collectingalgorithm, the two algorithms appeared in the thesis has a series of advantages, such as low performance loss, small load of collecting tasks and closely distance betweencollecting point and crucial moment. The thesis test task and application and proveRMS and EDF scheduling algorithm. Last but not least, promote the RTEMSapplicant scheduling algorithm approval and data collecting development oftechnology field.