When running OpenROAD on macro placements generated after AlphaChip training, I consistently encounter unresolved macro overlap issues.
In the paper, you mention a postprocessing step that legalizes macros by snapping them to the nearest power grid before fixing macro placements and running an EDA tool. This appears to address the overlap problem.
Could you share this script or provide details on how this legalization step is implemented?
When running OpenROAD on macro placements generated after AlphaChip training, I consistently encounter unresolved macro overlap issues.
In the paper, you mention a postprocessing step that legalizes macros by snapping them to the nearest power grid before fixing macro placements and running an EDA tool. This appears to address the overlap problem.
Could you share this script or provide details on how this legalization step is implemented?