Real time processing is embedded in an ever increasing number of systems. These systems integrate more and more functionalities which have to be activated in reaction to events coming from the system’s environment or from human users. Real time systems impose that the delay between an event and its reaction is bounded. These systems are generally programmed using a multi-tasking approach and tasks are handled at run time by a real time operating system (RTOS). The scheduler in the RTOS is in charge of allocating the CPU (e.g. a microcontroller) according to urgencies of the tasks which are generally represented as priority levels. In preemptive RTOS an execution of a task can be pre-empted as soon as a task with a higher priority is ready for execution. Each pre-emption results in a context switch which induces memory accesses for saving the context of the pre-empted task which will be restored back when the task will resume. Context switches imply also a pollution to the data and instruction caches which increases the context switch penalty. Therefore task pre-emption is a convenient mechanism to deal with real time applications but the overhead due to pre-emption can consume a part of the CPU time which is as many time in less for executing the tasks, and consumes energy as well.