Hardware-software co-design for a real-time executive

In the paper, we present hardware/software codesign methodologies for development of a real-time executive (RTEX). An RTEX is a special software that provides a minimum set of functions required for multitasking operation of embedded and real-time systems. In time critical environments, or where complex multitasking operations are employed, the execution time of such a software can prove a significant overhead on the system. This overhead may be significantly reduced by transferring some functions into a hardware co-processor-a software-in-silicon solution. We are using the RTEX as a prototyping tool for investigating different task scheduling schemes based on high performance evolutionary heuristics (e.g., genetic optimization and neurocomputing). In the paper, we present some of the results obtained by implementing a Hopfield neural network to solve the task scheduling problem in embedded and real-time systems. We compare these results with the performance of the well known scheduling policy Earliest Deadline First.