Abstract: The paper deals with the problem of linear decomposition of a system of Boolean functions. A novel analytic method for linearization, by reordering the values of the autocorrelation function, is presented. The computational complexity of the linearization procedure is reduced by performing calculations directly on a subset of autocorrelation values rather than by manipulating the Boolean function in its initial domain. It is proved that unlike other greedy methods, the new technique does not increase the implementation cost. That is, it provides linearized functions with a complexity that is not greater than the complexity of the initial Boolean functions. Experimental results over standard benchmarks and random Boolean functions demonstrate the efficiency of the proposed procedure in terms of the complexity measure and the execution time.
Keywords: Logic synthesis, spectral technique, autocorrelation function, linear transform, disjoint cubes.