We present a new algorithm for synthesis of reversible circuits for arbitrary n-bit bijective functions. This algorithm uses generalized Toffoli gates, which include positive and negative controls. Our algorithm is divided into two parts. First, we use partially controlled gen- eralized Toffoli gates, progressively increasing the number of controls. Second, exploring the properties of the representation of permutations in disjoint cycles, we apply generalized Toffoli gates with controls on all lines except for the target line. Therefore, new in the method is the fact that the obtained circuits use first low cost gates and consider increasing costs towards the end of the synthesis. In addition, we employ two bidirectional synthesis strategies to improve the gate count, which is the metric used to compare the results obtained by our algorithm with the results presented in the literature. Accordingly, our experimental results consider all 3-bit bijective functions and twenty widely used benchmark functions. The results obtained by our synthesis algorithm are competitive when compared with the best results known in the literature, considering as a complexity metric just the number of gates, as done by alternative best heuristics found in the literature. For example, for all 3-bit bijective functions using generalized Toffoli gates library, we obtained the best so far average count of 5.23.

One of the main motivations for reversible computing is that quantum computing has as one of its foundations the reversibility of all gates, that is, quantum computing circuit models are reversible. An important problem in reversible computing that has been intensively studied for the last decades is the synthesis of reversible circuits. The extended abstract considers optimization rules aiming to a new algorithm for post-synthesis optimization of reversible circuits composed of generalized Toffoli gates.