As AND-XOR network results in much better realization and requires fewer product terms than AND-OR realization, AND-XOR network has encouraged researchers to look for efficient minimization and synthesis tools for it’s realization. Among several canonical representations of AND-XOR networks, popular and the most testable one is the Fixed Polarity Reed Muller (FPRM) form. In this paper we have used GA (genetic algorithm) to select the polarities of the variables of the AND-XOR network. The polarities are selected based on the optimization of area, dynamic power and leakage power of the resulting circuit. This is the first ever effort to incorporate leakage power consideration in the variable polarity selection process. Here, we have presented new leakage power model of AND and XOR gates at 90nm technology. The area (in terms of number of product terms) results obtained are superior to those reported in the literature. The dynamic power minimization also results in their optimum values for many circuits that we have tested. It also enumerates the trade-offs present in the solution space for different weights associated with area, dynamic power and leakage power of the resulting AND-XOR network.

Fixed Polarity Reed Muller form, AND-XOR network, genetic algorithm, switching power, leakage power, Binary Decision Diagram.