Latch optimization in circuits generated from high-level descriptions

Abstract

In a gate-level description of a finite state machine (\fsm), there is a tradeoff between the number of latches and the size of the logic implementing the next-state and output functions. Typically, an initial implementation is generated via explicit state assignment or translation from a high-level language, and the tradeoff is subsequently only lightly explored. We efficiently explore good latch/logic tradeoffs for large designs generated from high-level specifications. We reduce the number of latches while controlling the logic size. We demonstrate the efficacy of our techniques on some large industrial examples.