Skip to main navigation Skip to search Skip to main content

Dr. Guide: AI-Guided Detailed Routing

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Routing is a crucial but complex stage in physical synthesis, where detailed routing (DR) is further known as the bottleneck for accelerating design cycles and improving circuit quality. Its difficulty lies mainly in the need to well organize numerous nets under various design rule constraints, leading to time-consuming iterative processes, e.g., rip-up and reroute (RRR), to achieve a convergent state. In this paper, we pioneer a new approach to assist DR with detailed AI guidance, where all nets are co-planned in advance to provide respective route guides to regulate the corresponding maze routing process. The proposed framework named Dr. Guide can be used as a flexible plug-in to existing routers and enjoys good interpretability. Incorporating it with one SOTA academic detailed router reveals its effectiveness and potential in accelerating/improving the DR process.

Original languageEnglish (US)
Title of host publication2025 ACM/IEEE 7th Symposium on Machine Learning for CAD, MLCAD 2025
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9798331537623
DOIs
StatePublished - 2025
Externally publishedYes
Event7th ACM/IEEE Symposium on Machine Learning for CAD, MLCAD 2025 - Santa Cruz, United States
Duration: Sep 8 2025 → Sep 10 2025

Publication series

Name2025 ACM/IEEE 7th Symposium on Machine Learning for CAD, MLCAD 2025

Conference

Conference7th ACM/IEEE Symposium on Machine Learning for CAD, MLCAD 2025
Country/TerritoryUnited States
CitySanta Cruz
Period9/8/25 → 9/10/25

Keywords

  • AI
  • generative model
  • physical design
  • routing

ASJC Scopus subject areas

  • Computer Graphics and Computer-Aided Design
  • Control and Optimization
  • Modeling and Simulation
  • Artificial Intelligence

Fingerprint

Dive into the research topics of 'Dr. Guide: AI-Guided Detailed Routing'. Together they form a unique fingerprint.

Cite this