State minimization of FSMs is a well-known problem . State minimization of completely specified FSMs (CSFSMs) has a complexity subquadratic in the number of states [50, 45]. This makes it an easy problem when the starting point is a two-level description of an FSM, because the number of states is usually less than a few hundred. The problem becomes difficult to manage when the starting point is an encoded sequential circuit with a large number of latches (in the hundreds). In that case the traditional method would be required to extract a state transition graph from the encoded network and then apply state minimization to it. But when latches are more than a dozen, the number of reachable states may be so huge to make state extraction and/or state minimization unfeasible. Recently it has been shown [90, 70, 73] how to bypass the extraction step and compute equivalence state pairs and equivalence classes of states implicitly. Equivalence classes are basically all that is needed to minimize a completely specified state machine. A compatible projection operator uniquely encodes each equivalence class by selecting a unique representative of the class to which a given state belongs. This implicit technique allows state minimization of sequential networks outside the domain of standard state minimization.
KeywordsClosure Condition State Minimization State Transition Graph Compatible Generation Closed Cover
Unable to display preview. Download preview PDF.