OpenFuILT-Eval

This repository contains the official code for full-chip scale OPC evaluation.

Install the OpenFuILT-Eval as a package using

python setup.py install

Dependency

To use this project, ensure the following dependencies are installed:

1. Environment: Python 3.11
2. PyTorch: With CUDA support for GPU acceleration
3. TorchLitho: Install from the official repository: TorchLitho-Lite
4. OpenFuILT: Install from the official repository: OpenFuILT

IO Specifications

• Pixel Size: The pixel of the mask represents an area of $s \times s \ nm^2$.
• Mask: A pickle file with a specified pixel size.
• Target: A GDSII file with a resolution of 1 nm.

Note: Ensure that the target and the mask are aligned at the source point $(0, 0)$. This implies:

​ The size of the mask is $[H, W]$, and the size of the target is $[H \times s, W \times s]$.

Support Metrics

• Edge Placement Error (EPE)
• L2 Losss
• Process Variation Band (PVB)
• Shot (Optional)

Example Code

The aligned inputs can be generated using OpenFuILT.

Below is an example code snippet demonstrating how to evaluate and obtain the metrics results:

from openFuILTEval import Evaluator

evaluator = Evaluator(
    pixel=14,
    macro_size=[2, 2],
    mask_path="mask.pkl",
    target_path="target.gds",
)

evaluator.evaluate()
evaluator.report()

You are also encouraged to use your generated full-chip mask and target GDSII files with the evaluation framework for comprehensive validation.

Citation

If you find this project helpful in your research, please consider citing our paper:

@inproceedings{yin2024fuilt,
  title={FuILT: Full chip ILT system with boundary healing},
  author={Yin, Shuo and Zhao, Wenqian and Xie, Li and Chen, Hong and Ma, Yuzhe and Ho, Tsung-Yi and Yu, Bei},
  booktitle={Proceedings of the 2024 International Symposium on Physical Design},
  pages={13--20},
  year={2024}
}