Desynchronisation Technique Using Petri Nets

Abstract

In this paper we consider the problem of desynchronising modular synchronous specifications for their realisation into GALS architectures and obtaining simple wrappers that are efficiently synthesisable using existing synthesis tools. The systems are modeled using Petri nets (PN) and the desynchronisation technique is based on the theory of PN Localities. The firing semantics of a globally synchronous system is characterised by maximal firing of input and output transitions. The partitioning of a synchronous system is achieved by unbundling the input transitions and allowing the output transitions to fire in maximal steps, in order to enable asynchronous communication in a distributed environment. Our model satisfies the two essential correctness properties, namely, semantics preservation and deadlock prevention, during the shift from maximal firing semantics, followed by synchronous systems, to standard interleaving semantics for input transitions and maximal step firing semantics for output transitions, followed by GALS architectures. The formation of localities is supported by adding internal signals which are necessary for building wrappers in the localities that will generate local clock enables. These wrappers can be subsequently synthesised using PN based synthesis tools.