Abstract:The design automation community has been actively exploring machine learning (ML) for very-large-scale-integrated (VLSI) computer-aided design (CAD). Many studies have explored learning-based techniques for cross-stage prediction tasks in the design flow. Although building ML models usually requires a large amount of data, most studies can only generate small internal datasets for validation due to the lack of large public datasets. Such a situation challenges the research in this field and raises potential issues like difficulty in benchmarking and reproducing results, limited research scope on small internal datasets, and high bar for new researchers. Therefore, in this article, we present an open-source dataset called “CircuitNet” for ML tasks in VLSI CAD. The dataset consists of more than 10K samples extracted from versatile runs of commercial design tools based on six open-source RISC-V designs which support typical cross-stage prediction tasks, such as routability and IR drop prediction, with extensive benchmarking on recent models. With the dataset prepared, we identify two practical challenges, data imbalance and model transferability, for ML application in CAD. To overcome data imbalance, we propose a loss function, biased loss, to give more weight to the minority, leading to 2% congestion reduction in routability-driven placement. We test the model transferability from RISC-V designs to ISPD 2015 contest designs in congestion prediction with several transfer learning methods and further proposed a knowledge distillation-based transfer learning framework with up to 20% accuracy improvement. We believe this dataset can open up new opportunities for ML in CAD research and beyond.

Implementation of Machine Learning In VLSI Global Routing

Global Routing Under a Congestion-Aware Reinforcement Learning Model

A Survey on Machine Learning-Based Routing for VLSI Physical Design

Machine Learning Based Routing Congestion Prediction in FPGA High-Level Synthesis

Implementation of Machine Learning in VLSI Integrated Circuit Design

Routing Congestion Prediction in VLSI Design Using Graph Neural Networks

An Efficient Global Router for Large-scale Congestion-driven Routing

A Machine Learning Approach to Routing

Reinforcement Learning based Interconnection Routing for Adaptive Traffic Optimization

Applying reinforcement learning to learn best net to rip and re-route in global routing

Machine Learning-Assisted Least Loaded Routing to Improve Performance of Circuit-Switched Networks

Towards Machine Learning for Placement and Routing in Chip Design: a Methodological Overview

AI/ML algorithms and applications in VLSI design and technology

Machine Learning-based Low Overhead Congestion Control Algorithm for Industrial NoCs

Prediction and Analysis of Machine Learning Models for Efficient Routing Protocol in VANET Using Feature Information

High-Speed Link Design Optimization Using Machine Learning SVR-AS Method

Accurate Prediction of Detailed Routing Congestion Using Supervised Data Learning.

Internet Intelligent Routing Architecture and Algorithm

Utilizing Machine Learning as a Prediction Scheme for Network Performance Metrics of Self-Clocked Congestion Control Algorithm

Machine Learning Based Traffic Prediction and Congestion Control Algorithms in Software Defined Networks

CircuitNet: an Open-Source Dataset for Machine Learning in VLSI CAD Applications with Improved Domain-Specific Evaluation Metric and Learning Strategies.