MaskPlace2.5D

Due to intellectual property restrictions, we are unable to directly provide the core code of MaskPlace2.5D. In the future, we will provide executable files for our leakage-aware thermal model to enable researchers to verify and extend it. However this repository contains a PyTorch implementation of the compact thermal model from the paper:

"ATPlace2.5D: Analytical Thermal-Aware Chiplet Placement Framework for Large-Scale 2.5D-IC"

The model provides fast thermal simulation for 2.5D integrated circuit designs with multiple chiplets. Thanks to ATPlace2.5D's open-source repository: https://github.com/Brilight/ATPlace_pub

🏗️ Architecture

Core Components

• compact_themal_model.py: Main PyTorch model implementation
• Thermal.py: HotSpot thermal solver interface
• process_thermal.py: Data processing and thermal simulation pipeline
• train_compact_themal_model.py: Training script for the compact model
• plot.py: Visualization utilities

📁 Project Structure

├── cases/                    # Test cases with chiplet layouts
│   ├── Case1/
│   │   ├── Case1_1.pl       # Layout file 1
│   │   ├── Case1_2.pl       # Layout file 2
│   │   ├── ...
│   │   ├── Case1.power      # Power consumption data
│   │   ├── Case1.intpsize   # Interposer size
│   │   └── Case1.blocks     # Block definitions
│   ├── Case2/
│   └── ...
├── model/                    # Trained model weights
│   ├── Case1_thermal_model.pth
│   ├── Case2_thermal_model.pth
│   └── ...
├── thermal/                  # HotSpot thermal simulator
│   ├── hotspot              # HotSpot executable
│   ├── hotspot.config       # Configuration file
│   └── new_hotspot.config   # Generated configuration
├── utils/                    # Utility functions
│   ├── blocks_parser.py     # Block definition parser
│   ├── fill_space.py        # Floorplan filler
│   ├── nets_parser.py       # Netlist parser
│   ├── pl_parser.py         # Placement file parser
│   └── uscs_parser.py       # UCS file parser
├── compact_themal_model.py  # Main thermal model
├── Thermal.py               # HotSpot interface
├── process_thermal.py       # Data processing
├── train_compact_themal_model.py  # Training script
├── train_compact_themal_model.sh  # Batch training script
└── README.md               # This file

🚀 Quick Start

Prerequisites

• Python 3.10+
• PyTorch 1.8+
• NumPy
• HotSpot thermal simulator (included in thermal/ directory)
• CUDA-enabled GPU (recommended)

Installation

1. Clone this repository:

git clone https://github.com/xunk33333/MaskPlace2.5D

2. Install required Python packages:

pip install torch numpy tqdm

Prepare Dataset

python porcess_thermal.py

Training the Model

To train the compact thermal model for a specific case:

python train_compact_themal_model.py --case 1 --train

This will:

1. Train the compact thermal model
2. Save the trained model to model/Case1_thermal_model.pth
3. Calculate performance metrics (MAE, RMSE, etc.) and generate comparison plots

Batch Training

To train models for all 10 cases simultaneously:

bash train_compact_themal_model.sh

Testing the Model

To only test a trained model:

python train_compact_themal_model.py --case 1

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🆘 Support

For questions and support, please open an issue in the GitHub repository.