Got different pin locations to die boundaries

[young03600]

Hi, I would like to know how the pin coordinates are extracted using dreamplace?

Thanks.

Hi, you can read the DEF into DREAMPlace, and call the pinpos operator to get a tensor of pin positions. You can read DREAMPlace's HPWL operator to see how its inputs are constructed (their input contains the pin coordinate tensor).
Originally posted by @gzz2000 in #8 (comment)

Hi,

I previously posted a related inquiry in another thread, but since that thread has been marked as completed, I believe it would be better to post this as a new issue.

As stated in the title, I’m seeing discrepancies in pin locations when using the pin_pos operation mentioned in that issue. I retrieved the pin locations for comparison, but they don’t exactly match the locations given by the nf[2:5] dump. (For reference, I used the “spm” design for testing.)

Additionally, I reviewed the description of nf[2:5] in the repository. The pin order doesn’t appear to follow the expected top/left/right/bottom layout of the die area. It seems that nf[2] + nf[5] and nf[3] + nf[4] do not match the die area height and width, respectively.

Could you please clarify this discrepancy?

The following is the minimum code to reproduce.

import sys
import os
import logging
# for consistency between python2 and python3
root_dir = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
if root_dir not in sys.path:
    sys.path.append(root_dir)
import numpy as np
import torch
import PlaceDB
import Params
from dreamplace.ops.pin_pos.pin_pos import PinPosFunction, PinPosSegmentFunction
from dreamplace.ops.hpwl.hpwl import HPWLFunction, HPWLAtomicFunction
from dreamplace.ops.pin_pos.pin_pos import PinPos
from dreamplace.ops.hpwl.hpwl import HPWL

if __name__ == '__main__':

    np.set_printoptions(threshold=np.inf, suppress=True)

    params = Params.Params()
    params.load(sys.argv[1])  # only read the DEF/LEF file in this json file
    logging.info("parameters = %s" % (params)) 
    placedb = PlaceDB.PlaceDB()
    placedb(params)

    pin_offset_x = torch.from_numpy(placedb.pin_offset_x).to(torch.float).cuda()
    pin_offset_y = torch.from_numpy(placedb.pin_offset_y).to(torch.float).cuda()
    pin2node_map = torch.from_numpy(placedb.pin2node_map).to(torch.long).cuda()
    flat_node2pin_map = torch.from_numpy(placedb.flat_node2pin_map).to(torch.int).cuda()
    flat_node2pin_start_map = torch.from_numpy(placedb.flat_node2pin_start_map).to(torch.int).cuda()

    pin_pos_instance = PinPos(
        pin_offset_x=pin_offset_x,
        pin_offset_y=pin_offset_y,
        pin2node_map=pin2node_map,
        flat_node2pin_map=flat_node2pin_map,
        flat_node2pin_start_map=flat_node2pin_start_map,
        num_physical_nodes=placedb.num_physical_nodes,
    )
    pin_pos_instance = pin_pos_instance.cuda()
    pos = torch.cat([torch.from_numpy(placedb.node_x).to(torch.float), 
                     torch.from_numpy(placedb.node_y).to(torch.float)]).cuda()
    pin_pos = pin_pos_instance(pos)
    
    # only use for checking the pin locations
    pin_pos = np.array(pin_pos.cpu())
    pin_x = pin_pos[:placedb.num_pins]
    pin_y = pin_pos[placedb.num_pins:]
    pin_pos = np.stack((pin_x, pin_y), axis=1)

    left = placedb.xl
    right = placedb.xh
    bottom = placedb.yl
    top = placedb.yh

    relative_distances = np.array([
        pin_pos[:, 0] - left,      # distance_to_left
        right - pin_pos[:, 0],     # distance_to_right
        pin_pos[:, 1] - bottom,    # distance_to_bottom
        top - pin_pos[:, 1],       # distance_to_top
    ]).T
    relative_distances = np.array(relative_distances)
    with open('pins', mode='w+', encoding='UTF-8') as file:
        file.write(str(relative_distances))
        logging.info('write successfully')

Thanks in advance for your help.

[gzz2000]

Hi, thanks for your question! Can you try [2]+[3] and [4]+[5]? We made a mistake in the description and the order should be top/bottom/left/right..
Also, the die width and height are in the unit of site width.

[young03600]

Hi, thanks for your question! Can you try [2]+[3] and [4]+[5]? We made a mistake in the description and the order should be top/bottom/left/right.. Also, the die width and height are in the unit of site width.

Hi,
Thanks for the reply. I found the correct order should be left/right/bottom/top accroding to the dumped 'nf' from graph.bin
and the placedb seems to automatically shift and scale the coordinate. Does it possible to provide a snippet of the json file read by the placedb?

Could you please help me to clerify that I'm using the correct cell positions to get the correct pin locations?
According to the provided snippet, I found that only the bottom/top values got different from the dumped values from graph.bin

[Rahelehabbasii]

Could you please share the full dataset or link for download?

[gzz2000]

Could you please share the full dataset or link for download?

There are links on readme page https://github.com/TimingPredict/Dataset?tab=readme-ov-file