# EDALearn: A Comprehensive RTL-to-Signoff EDA Benchmark for Democratized and Reproducible ML for EDA Research

**Authors:** Jingyu Pan, Chen-Chia Chang, Zhiyao Xie, Yiran Chen  
**arXiv:** [2312.01674](https://arxiv.org/abs/2312.01674)  
**Date:** 2023-12-04  
**Summary page:** https://www.summarizepaper.com/en/paper/2312.01674v1/

## Key points

- Integration of machine learning techniques is a focal point in the realm of electronic design automation (EDA).
- Limited access to comprehensive public datasets has led to the use of smaller, internally generated datasets in EDA research.
- EDALearn is an innovative open-source initiative tailored for ML tasks within EDA, covering the entire spectrum from RTL synthesis to signoff for physical implementation.
- EDALearn aims to democratize and promote reproducibility in ML research for EDA applications by enriching data collection across multiple stages of VLSI design processes.
- The benchmark suite within EDALearn facilitates a deeper understanding of ML transferability across diverse technology nodes and accommodates a wide range of VLSI design instances and sizes.
- Comprehensive data analysis capabilities in EDALearn provide detailed insights into dataset attributes and distribution, enabling users to develop highly efficient ML models for specific EDA challenges.
- EDALearn aims to drive progress in VLSI design methodologies by fostering innovation at the intersection of machine learning and EDA domains with an emphasis on openness, reproducibility, and inclusivity.

## Summary

In the realm of  for , the integration of  techniques has become a focal point of research. The effectiveness of ML models in this domain relies heavily on the availability of extensive datasets, yet many studies are constrained by limited access to comprehensive public resources, leading to the use of smaller, internally generated datasets. To address this gap, a groundbreaking initiative known as  has been introduced. This innovative open-source  is specifically tailored for ML tasks within EDA, offering a holistic approach that covers the entire spectrum from RTL synthesis to signoff for physical implementation. EDALearn stands out for its commitment to democratizing and promoting reproducibility in ML research for EDA applications. By enriching data collection across multiple stages of VLSI design processes, this benchmark suite facilitates a deeper understanding of ML transferability across diverse technology nodes. Moreover, EDALearn accommodates a wide range of VLSI design instances and sizes, accurately reflecting the complexity inherent in modern VLSI designs. One key feature that sets EDALearn apart is its comprehensive data analysis capabilities. Users are provided with detailed insights into the attributes and distribution of the dataset, empowering them to develop highly efficient ML models tailored to specific EDA challenges. Through these contributions, EDALearn aims to catalyze further advancements in the intersection of ML and EDA domains, fostering innovation and driving progress in VLSI design methodologies. With its emphasis on openness, reproducibility, and inclusivity, EDALearn represents a significant step forward in advancing research at the cutting edge of electronic design automation.