On the benefits of kernel extraction during logic optimization

Kernel extraction during logic minimization, although beneficial in most cases, may not be applicable to certain classes of functions due to the restructuring effects this operation may have on the Boolean network being optimized. In this paper, we identify and characterize a class of logic networks for which “blind” kernel extraction does not pay off in terms of both power and area minimization. More specifically, we provide experimental evidence that arithmetic operations are negatively affected by kernel extraction