System, method, and computer-program product for routing in an electronic design using deep learning

A non-transitory computer-readable storage medium for routing in an electronic design, the computer-readable storage medium having stored thereon instructions that when executed by a machine result in one or more operations, the operations comprising:

• receiving, using at least one processor, global route data associated with the electronic design as an input;
• generating detail route data, based upon the global route data, wherein the global route data includes a pin location shape, a standard pitch of a design, a pin-to-pin connection, a blockage location, a blockage shape, and a pre-routed net;
• receiving at least one of a post-route physical layout information and a pre-route physical layout information, wherein the at least one of the post-route physical layout information and the pre-route physical layout information includes a routing wire and a power/ground net;
• transforming one or more of the detail route data, the global route data, the post-route physical layout information, and the pre-route physical layout information into one or more numerical vectors including at least one input feature and at least one output result of a deep neural network;
• training, based upon at least in part, the one or more numerical vectors, the deep neural network with one or more of the global route data, the detail route data, the post-route physical layout information, and the pre-route physical layout information, wherein the neural network is configured to approximate a mathematic function corresponding to a detail routing algorithm;
• and
• predicting post-route resistance and capacitance (“RC”) parasitic parameter and congestion information prior to routing, wherein predicting includes predicting an output associated with a detail route based upon, at least in part, a trained deep neural network model, wherein predicting an output includes predicting a wire length or a via number for at least one net associated with the electronic design.

A system for routing in an electronic design comprising:

• a computing device having at least one processor configured to receive global route data associated with the electronic design as an input, the at least one processor further configured to generate detail route data, based upon the global route data, wherein the global route data includes a pin location shape, a standard pitch of a design, a pin-to-pin connection, and a pre-routed net, the at least one processor further configured to receive at least one of a post-route physical layout information and a pre-route physical layout information, wherein the at least one of the post-route physical layout information and the pre-route physical layout information includes at least one of a netlist, a cell location, a routing wire, and a power/ground net, the at least one processor further configured to transform one or more of the detail route data, the global route data, the post-route physical layout information, and the pre-route physical layout information into one or more numerical vectors including at least one input feature and at least one output result of a deep neural network, the at least one processor further configured to train, based upon at least in part, the one or more numerical vectors, the deep neural network with the global route data and the detail route data, the at least one processor further configured to predict post-route resistance and capacitance (“RC”) parasitic parameter and congestion information prior to routing, wherein predicting includes predicting an output associated with a detail route based upon, at least in part, a trained deep neural network model, wherein predicting an output includes predicting a wire length and or a via number for at least one net associated with the electronic design.