A Model of Parallel Deterministic Real-Time Computation

This paper presents a model of computation based on real-time constraints and asynchronous message passing, and proves a sufficient and necessary condition for this model to be deterministic. The model is then extended with deterministic error handling, meaning that the same error yields the same consequences on the system. We consider two different error occurrence models: at a specific time, or at a specific instruction, and conclude that the "error at a specific time" model is more suitable for practical use. We proceed by presenting a concrete implementation of this model in the PharOS real-time system.