Using fuzzy logic to derive an optimum embedded application specific architecture

Embedded computer system design is traditionally ad hoc and manual. In most instances, system designers put together a “straw-man” system and then evolve from there using classical queue network model (QNM). This approach can cause tunnel vision, since system designers tend to put together a system that is very similar to the predecessor system. The resulting architecture is not derived directly from the set of target application code, instead it is only a modification of the predecessor system. This paper will introduce the concept behind the architecture process flow that uses compiler front-end technique, knowledge base, neural nets and fuzzy logic rules to derive an application specific architecture directly from the target application requirements. A case study on deriving an architecture with limited set of components from the Motorola M68HC11 microcontroller family will be demonstrated using the architecture process flow