Finite State Machine Decomposition

Abstract

Finite state machine (FSM) decomposition is concerned with the implementation of a FSM as a set of smaller interacting submachines. Such an implementation is desirable for a number of reasons. A partitioned sequential circuit usually leads to improved performance as a result of a reduction in the longest path between latch inputs and outputs. This fact is particularly true when the individual submachines are implemented as Programmable Logic Arrays (PLAs). It appears that the primary interest in using decomposition tools in industry stems from a need to improve the performance of FSM controllers, which often dictates the required duration of the system clock. FSM decomposition can be applied directly when Programmable Gate Array (PGAs) or Programmable Logic Devices (PLDs) are the target technology. Such technologies are characterized by I/O or gate-limited blocks of logic and latches into which the circuit must be mapped. In many cases, it is desirable for reasons of clock-skew minimization or simplifying the layout to distribute the control logic for a data path in such a manner that the portions of the data path and control that interact closely are placed next to each other. FSM decomposition can also be used for this purpose. Partitioning of the logic implementing the FSM could result in simplified layout constraints resulting in smaller chip area. In PLA-based FSMs, decomposition has the effect of partitioning the PLA that implements the original FSM into smaller interacting PLAs that implement the individual submachines. In such situations, an area reduction can be attributed to PLA partitioning. Finally, it is not computationally feasible for current multilevel logic minimizers (e.g. MIS-II [10]) to search all possible area minimal solutions. In some cases, an initially-decomposed FSM could correspond to a superior starting point for multilevel logic minimization.