Initializability Analysis of Synchronous Sequential Circuits

This paper addresses the problem of initializing synchronous sequential circuits, i.e., of generating the shortest sequence able to drive the circuit to a known state, no matter the initial state. Logic initialization is considered, being the only one compatible with current commercial tools. A Hybrid Genetic Algorithm is proposed, which combines general ideas from evolutionary computation with specific techniques, well suited to the addressed problem. For the first time, experimental results provide data about the complete set of ISCAS'89 circuits, and show that, despite the inherent algorithm incompleteness, the method is capable of finding the optimum result for the considered circuits. A prototypical tool implementing the algorithm found better results than previous methods.