accel-stack

KO | EN

gitlite — search

#javascript #python #hacktoberfest #react #ai #typescript #llm #go #golang #android #machine-learning #rust #deep-learning #linux

★ 11 Open GitHub ↗

AccelStack: A Cost-Driven Analysis of 3D-Stacked LLM Accelerators

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# AccelStack

*Chen Bai, Xin Fan, Zhenhua Zhu, Wei Zhang, Yuan Xie. AccelStack: A Cost-Driven Analysis of 3D-Stacked LLM Accelerators. In Proceedings of the ACM/IEEE International Conference on Computer-Aided Design (ICCAD), 2025.*

This repo currently only includes code related to cost analysis. The code is based on *Chiplet Actuary*, which is a quantitative cost model for multi-chiplet systems.

*Yinxiao Feng and Kaisheng Ma. Chiplet Actuary: A Quantitative Cost Model and Multi-Chiplet Architecture Exploration. In Proceedings of the 59th ACM/IEEE Design Automation Conference (DAC), 2022.*

## Project Structure

- `main.py`: Main entry point with cost analysis functions
- `exploration.py`: Additional analysis functions
- `parameter.ini`: Configuration file for model parameters
- `cost_model/`: Core library implementing the chiplet cost model

## Features

- Cost analysis for monolithic, 3D, and 3.5D architectures
- Support for different packaging technologies (OS, CoWoS, EMIB)
- Configurable parameters through `parameter.ini`
- Volume-based cost sweeping analysis

## Usage

### Requirements

- Python 3.12
- Pandas 2.2

### Quick Start

1. Configure parameters in `parameter.ini` as needed
2. Run the main analysis:
   ```bash
   python main.py
   ```

### Configuration

The `parameter.ini` file contains all configurable parameters for the cost model. 
You can modify this file to adjust technology parameters, manufacturing costs, and other factors.
Currently, numbers in `parameter.ini` are referenced to ICKnowledge.

*"IC Knowledge LLC. IC Cost and Price Model; Assembly and Test Cost and Price Model." 2021. https://www.techinsights.com/press-release/techinsights-acquires-ic-knowledge-llc.*

To use a custom parameter file location, use the `set_parameter_path()` function:
```python
from cost_model import spec
spec.set_parameter_path("/path/to/your/parameter.ini")
```

### Functions

- `generate_cost_volume_sweep()`: Generates cost analysis across different volumes and saves to Excel

## Authors

- eecbai@ust.hk
- xfanat@connect.ust.hk

accel-stack

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

accel-stack

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README