Dynamic control of reliable execution of parallel computing processes for real-time systems

The proposed method of control of parallel computing processes relies on static forecasting of the execution times of involved sets of interrelated tasks and/or their fragments (program modules) in parallel computer systems, as well as on dynamic updating of static forecasts daring the task execution: the actual time parameters of ea;executing some program modules are compared with the updated forecasts of the same parameters. The results of comparison are used to choose the process control algorithm (from a given set of algorithms) that would ensure the reliable execution of the set of interrelated tasks, that is, their execution in a time not exceeding the prescribed time with the required probability. Control of reliable execution of the tasks with logical branchings between the program modules is discussed for the first time.