An Efficient Global Router for Large-scale Congestion-driven Routing

In the very large scale integration (VLSI) global routing stage, one of the most critical considerations is how to obtain a proper balance between routing quality and runtime. In this paper, we adopt a multi-threading technique to accelerate various subphases of global routing. To ensure the robustness of routing results, we add a stage called congestion and wire length aware breaking circles algorithm while ensuring to remove the most congested or longest wire segment in the loops. Since routing congestion is the most critical constraint in the global routing phase, we present two accurate cost models to evaluate the routing resources, which can effectively reduce the congestion of routing results. The experiment shows that our proposed method reduces the violations by 28.3% and achieves a 2x faster runtime in benchmarks of the ISPD’19 contest compared with CUGR.