A computer organisation suitable to improve the performance of embedded systems encountered in hard real-time environments is described, which enhances dependability and efficiency by exploiting possibilities for parallel processing inherent to real-time systems. It is shown that internal data transmissions due to context-switching can be eliminated and that CPUs can be relieved from a considerable amount of routine work by providing a separate module for major functions of real-time operating systems, viz., for interrupt and signal recognition, for time management, and for the administration of task execution plans and states. Considering the layer model of contemporary real-time operating systems, these functions constitute kernel and first layer. The hardware implementation of typical hard real-time support features provides a clear physical separation of the intrinsically independent functions event reaction and (user) task processing. While minimising response times in general, the architecture enables to guarantee predefined, short upper bounds for them. It also incorporates a new approach for and hardware realisation of timers characterised by utmost precision and high handling comfort. Such timers are particularly well-suited to serve in distributed real-time systems, since, by employing the exact UTC received from the satellites of GPS (the global navigation and positioning system), synchronisation is not needed anymore.
