Official resources of "BYOS: Knowledge-driven Large Language Models Bring Your Own Operating System More Excellent ". Hongyu Lin, Yuchen Li, Haoran Luo, Kaichun Yao, Libo Zhang, Mingjie Xing, Yanjun Wu. [paper]

BYOS (Bring Your Own Operating System more excellent) is a novel knowledge-driven framework that automates Linux kernel tuning using Large Language Models (LLMs). BYOS achieves 7.1%-155.4% performance improvements over default configurations across standard benchmarks and real-world applications.

• 🏗️ Structured Knowledge Construction: Dual-layer Knowledge Graph (OD-KG) bridges high-level tuning objectives with low-level config options.
• 🧠 Knowledge-driven Configuration: Mitigates LLM hallucinations through graph-guided reasoning.
• 🔄 Continuous Knowledge Maintenance: Incremental updates adapt to rapid kernel evolution.
• 🚀 Performance Gains: Up to 155.4% improvement on UnixBench, 42.7% latency reduction in Nginx.

git clone https://github.com/LHY-24/BYOS.git
cd BYOS
pip install -r requirements.txt

1. Init environment

# replace your OpenAI key (and base_url) in Config.py file and init_env.sh
source init_env.sh
# downloads linux kernel source code from https://www.kernel.org/pub/linux/kernel/. Use v6.2.16 as an example.
wget https://www.kernel.org/pub/linux/kernel/v6.x/linux-6.2.16.tar.gz
tar -zxf https://www.kernel.org/pub/linux/kernel/v6.x/linux-6.2.16.tar.gz

2. Initialize the knowledge base:

python3 build_kg.py /path/to/linux/kernel/source_code

3. Run kernel tuning:

python3 BYOS.py /path/to/linux/kernel/source_code -t "Your optimize target" -o "config output filename"

4. Apply configurations:

Move the config generated by Step 3 to linux source code directory and rename it to ".config". Then compile the linux kernel.

mv config_output /path/to/linux/kernel/source_code/.config

@misc{lin2025byosknowledgedrivenlargelanguage,
      title={BYOS: Knowledge-driven Large Language Models Bring Your Own Operating System More Excellent}, 
      author={Hongyu Lin and Yuchen Li and Haoran Luo and Kaichun Yao and Libo Zhang and Mingjie Xing and Yanjun Wu},  
      year={2025},
      eprint={2503.09663},
      archivePrefix={arXiv},
      primaryClass={cs.OS},
      url={https://arxiv.org/abs/2503.09663}, 
}

Contributions and feedback are greatly appreciated! Whether you've found a bug, have a question, or want to suggest improvements, please open an issue. Your input helps make BYOS better for everyone.

For further questions, please contact: hongyu2021@iscas.ac.cn, liyuchen2021@iscas.ac.cn, luohaoran@bupt.edu.cn.

This repo benefits from AutoOS and LightRAG. Thanks for their wonderful works.