Dynamic Scheduling of Imprecise Hard Real-Time Tasks

Imprecise computation can reduce the proportion of tasks violating their time constraints in a hard real-time system. This is achieved by executing tasks for a reduced amount of time, but at the cost of returning imprecise results. This paper discusses dynamic scheduling algorithms for determining how much time should be allocated to each imprecise task. The performance results show that these algorithms are shown to be effective at trading off the preciseness of results for the task loss. 1 Introduction In hard real-time computer systems, tasks have time constraints such as deadlines. Unless these time constraints are satisfied, the computational results of a task are of little value and the task is said to be lost. Consequently, a major goal in the operation of hard real-time computer systems is to minimize the proportion of tasks that are lost. One way to do this is to use the imprecise computation technique. In imprecise computation, a task is allowed to return imprecise results i..