Clan-Based Structural Circuit Decomposition for Template Generation

We describe in this paper a novel approach to identify repeating circuit structures referred to as templates. The approach is based on a clan-based decomposition algorithm that has been successfully used in other domains 6, 5]. The algorithm discovers clans, a grouping of nodes in the circuit graph that have a natural aan-ity towards each other, and it hierarchically builds a parse tree of all clans. A non-leaf node in the parse tree corresponds to a clan. Each clan is classiied as either linear, independent, or pseudo-linear to reeect the relationships among the node's immediate children, which are either clans or leaf nodes. Identical clans in the parse tree are then identiied to be appropriate templates with which the circuit can be covered. We describe the results of applying this algorithm to a ripple adder, an ALU, and 4 4 multiplier.