Reducing the Maximum Length of Connections in Single Flux Quantum Circuits During Routing

As the number of nets connecting single-flux-quantum (SFQ) cells in large-scale SFQ circuits grows, powerful electronic design automation (EDA) tools are needed to mitigate the wire routing task. Moreover, the clock frequency of SFQ circuits is heavily influenced by the longest wire delay. However, current routing tools have no means to control the maximum length of routing wires. In this paper, we present an innovative post-routing optimization framework which reduces the maximum wirelength in SFQ circuits. A framework is developed in which the longest wire is ripped and re-routed by resorting to a maze routing algorithm after the acquisition of wire density distribution using a machine learning method. Based on the MIT-LL SFQ5ee process technology and using a small library of SFQ logic cells, we show that the proposed framework can complete post-routing optimization of 13 SFQ circuits in 8 minutes while reducing the length of the longest wire by 11.8% on average over the state-of-the-art EDA routing tool for large-scale SFQ circuits.