IIT Guwahati and NYU unveil ‘LEAP’, a groundbreaking ML framework set to revolutionise IC design with unprecedented efficiency and precision in the field of EDA.
Researchers from the Indian Institute of Technology Guwahati (IIT-G), in collaboration with New York University (NYU), have launched a pioneering machine learning framework called ‘LEAP’ that is set to transform Electronic Design Automation (EDA). This innovative tool enhances Integrated Circuit (IC) design—a critical element of the $600 billion semiconductor industry.
Developed by Prof. Chandan Karfa, Dr. Sukanta Bhattacharjee, their BTech students, and NYU’s Dr. Animesh Basak Chowdhury, LEAP offers a substantial leap forward in design efficiency.
The framework employs advanced machine learning techniques to intelligently narrow potential design configurations, cutting the number of options considered by over 50%. This optimisation streamlines the technology mapping process, making it significantly more efficient.
In tests conducted on 21 different designs, LEAP demonstrated a 50% improvement in runtime and reduced the number of configurations checked by over 51%.
Additionally, LEAP achieved a 2% reduction in the clock period without increasing the circuit area, boosting the performance of the open-source ABC EDA tool by 63%.
Traditional IC design often relies on heuristic methods, which offer quick but sometimes suboptimal solutions. LEAP’s advanced approach addresses these challenges by enhancing design quality and efficiency, representing a significant advancement in electronic design automation.
Prof. Karfa emphasised that LEAP’s circuit performance and design time improvements are essential milestones in the field.
