Static analysis techniques for the synthesis of efficient asynchronous circuits

In the context of deriving asynchronous circuits from high-level descriptions, determining whether two actions are potentially concurrent (overlapped execution) or serial (non-overlapped execution) has several advantages. This knowledge can be utilized to efficiently implement shared variables, support speculative guard evaluation, and optimize resources (circuitry) by sharing. In a distributed environment with several concurrent processes, determining whether two actions are potentially concurrent or not, automatically, is often difficult to formulate and computationally expensive. In this paper, we present techniques to overcome these problems. First, we present a tool called parComp which infers the composite behavior of a collection of modules, and then we present an algorithm called conCur to analyze the inferred behavior to detect the seriality of two actions. Simple heuristics are presented for the abstraction of the inferred behavioral descriptions and improving the efficiency of conCur. The algorithms parComp and conCur are illustrated in the hopCP framework and implemented in Standard ML of New Jersey. Execution times of the algorithms are reported on a variety of examples. The results are quite encouraging.