CSE291_Project#2

Script	Purpose
netlist2graph.py	Transform the netlist to a hypergraph
loc2def.py	Read the predicted initial locations back into def file.

---

1. netlist2graph.py

This script parsers the netlist to a hypergraph

• Inputs:
  • 3_2_place_iop.def generated by OpenROAD
• Outputs:

  • hypergraph.txt

    Number of vertices: <total>
      Number of macros + std_cells: <total_cell_number>
      Number of IOs: <total_io_number>
    hyperedges: driver_id load_id1 load_id2 ...
    0 17 25
    1 4 9 11
    ...
    

  • vertex_info.txt

    vertex_id, vertex_name, is_fixed, x, y
    0, U0, 0, 0, 0
    1, U1, 1, 2280, 2800
    2, PAD0, 1, 45500, 0
    ...
    

2. loc2def.py

This script updates the DEF file (3_2_place_iop.def) by embedding predicted locations for movable cells.

• Inputs:

  • A CSV file containing (x, y) coordinate pairs for all movable cells.

    • [Note] Coordinates include only movable cells, and their order must matches the order of components listed in the COMPONENTS section.

  • The original 3_2_place_iop.def

    - _2975_ INV_X1 + UNPLACED ;
    - _2976_ BUF_X1 + UNPLACED ;
    - _2977_ INV_X1 + UNPLACED ;
    - _2978_ BUF_X1 + UNPLACED ;
    

• Outputs:

  • Modified 3_2_place_iop.def

    - _2975_ INV_X1 + PLACED ( 1192114 2709123 ) N ;
    - _2976_ BUF_X1 + PLACED ( 1378472 2889836 ) N ;
    - _2977_ INV_X1 + PLACED ( 1376774 2889722 ) N ;
    - _2978_ BUF_X1 + PLACED ( 1375933 2889801 ) N ;
    

How to use them?

• Use OpenROAD to generate the 3_2_place_iop.def.

• Use netlist2graph.py to convert the .def file into a hypergraph format.

• Extract necessary features (you could modify or write your own parsers to extract additional features if needed).

• Train models based on these features.

• Use loc2def.py to write the predicted cell locations back into the .def file.

• Replace the original .def in OpenROAD and run global_placement -skip_initial_place -incremental to continue the flow and obtain the detailed placement results