Differentiable OPC

The developing framework is beased on pytorch + hydra configuring template. FRAMEWORK README

Paper

Differentiable Edge-based OPC Guojin Chen¹ (work done during internship at NVIDIA Research), Haoyu Yang², Haoxing Mark Ren², Bei Yu¹, David Z. Pan³. ¹The Chinese University of Hong Kong · ²NVIDIA Research, Austin, USA · ³ECE, The University of Texas at Austin. Proc. 43rd IEEE/ACM International Conference on Computer-Aided Design (ICCAD '24), pp. 1–9, ACM, 2024. DOI: 10.1145/3676536.3676764

If you use this code, please cite:

@inproceedings{chen2024diffopc,
  author    = {Chen, Guojin and Yang, Haoyu and Ren, Haoxing Mark and Yu, Bei and Pan, David Z.},
  title     = {Differentiable Edge-based {OPC}},
  booktitle = {Proceedings of the 43rd IEEE/ACM International Conference on Computer-Aided Design (ICCAD)},
  year      = {2024},
  pages     = {1--9},
  publisher = {ACM},
  address   = {New York, NY, USA},
  doi       = {10.1145/3676536.3676764},
  url       = {https://dl.acm.org/doi/10.1145/3676536.3676764},
}

🚀  Quickstart

# clone project
git clone https://github.com/dekura/DiffOPC.git
cd DiffOPC

# [OPTIONAL] create conda environment
conda create -n dopc python=3.10
conda activate dopc

# install pytorch according to instructions
# https://pytorch.org/get-started/

# install requirements
pip install -r requirements.txt

Run DiffOPC

Run a single ICCAD-2013 test case (debug config, 70 ILT iterations):

python src/diffopc.py opc=debug data=single data.data_idx=1 extras.print_config=false

Run the full 10-case ICCAD-2013 sweep:

python src/diffopc.py opc=debug data=default extras.print_config=false

Disable the Aim logger (skip writing to ./aim/):

python src/diffopc.py opc=debug data=single data.data_idx=1 extras.print_config=false logger=[]

Reference results (debug config, single GPU)

Verified on torch 2.11 / CUDA 13, single-case smoke runs:

Case	L2	PVBand	EPE	BestIter	Runtime
1	41,159	60,955	3	59	19.2 s
3	68,771	103,848	24	69	30.6 s

Notes

• Use opc=debug rather than opc=default — the default config is missing SRAF_FORBIDDEN and IsInsertSRAF, which are referenced by the data loader and the solver.
• Shot counting (eval_shot=True) requires the adabox package, which is not on PyPI. The debug config has eval_shot: False, so this is only relevant if you turn shot counting on.

Project Structure

The directory structure:

├── .github                   <- Github Actions workflows
│
├── configs                   <- Hydra configs
│   ├── data                     <- Data configs: ICCAD or larger dataset
│   ├── experiment               <- Experiment configs
│   ├── extras                   <- Extra utilities configs
│   ├── hydra                    <- Hydra configs
│   ├── logger                   <- Logger configs : aim stack logger
│   ├── opc                      <- OPC model configs
│   ├── litho                    <- litho model configs
│   ├── sraf                     <- sraf model configs
│   ├── paths                    <- Project paths configs
│   ├── solver                   <- Solver configs
│   │
│   ├── multidiff.yaml           <- Multilevel config for optimization
│   └── diffopc.yaml             <- Main config for optimization
│
├── benchmark                   <- Project data
│
├── logs                   <- Logs generated by hydra and lightning loggers
│
├── scripts                <- Shell scripts
│
├── src                    <- Source code
│   ├── data                     <- Data scripts
│   ├── opc                      <- OPC Model scripts
│   ├── utils                    <- Utility scripts
│   │
│   └── diffopc.py                 <- Run training
│
├── tests                  <- Tests of any kind
│
├── .env.example              <- Example of file for storing private environment variables
├── .gitignore                <- List of files ignored by git
├── .pre-commit-config.yaml   <- Configuration of pre-commit hooks for code formatting
├── .project-root             <- File for inferring the position of project root directory
├── environment.yaml          <- File for installing conda environment
├── Makefile                  <- Makefile with commands like `make train` or `make test`
├── pyproject.toml            <- Configuration options for testing and linting
├── requirements.txt          <- File for installing python dependencies
├── setup.py                  <- File for installing project as a package
└── README.md