Detailed routing violation prediction during placement using machine learning

The complexity of design rules at 22nm and below precludes direct incorporation of detailed routing (DR) rules into a placement algorithm. However, ignoring routability rules during the placement process may result in infeasible designs. The congestion estimated by a global router is conventionally used for routing estimation during placement, but it does not include real detailed routing violations, which adversely affect the routability of a design. Presently, there are no methods that directly aim to predict detailed routing violations. In this paper we propose a machine learning based method to predict the shorts that are a major component of detailed routing violations. The proposed method can be integrated into a placement tool and be used as a guide during the placement process to reduce the number of shorts happening in the detailed routing stage. Empirical results show that our method is successful in predicting 88% of the shorts with only 16% incorrectly predicting shorts in no short violation area.