The automatic synthesis of low power finite-state machines (FSM's) with gated clocks relies on efficient algorithms for synthesis and optimization of dedicated clock-stopping circuitry. We describe a new transformation for incompletely specified Mealy-type machines that makes them suitable for gated-clock implementation with a limited increase in complexity. The transformation is probabilistic-driven, and identifies highly-probable idle conditions that will be exploited for the optimal synthesis of the logic block that controls the local clock of the FSM. We formulate and solve a new logic optimization problem, namely, the synthesis of a subfunction of a Boolean function that is minimal in size under a constraint on its probability to be true. We describe the relevance of this problem for the optimal synthesis of gated clocks. A prototype tool has been implemented and its performance, although influenced by the initial structure of the FSM, shows that sizable power reductions can be obtained using our technique.
展开▼
