IIT Guwahati's New ML Framework 'LEAP' Revolutionizes Integrated Circuit Design

Researchers at IIT Guwahati have developed 'LEAP', a Machine Learning framework transforming Electronic Design Automation. It enhances Integrated Circuit designs, crucial for the $600 billion semiconductor industry, by improving efficiency and reducing runtime by 50%. This innovation promises faster, more energy-efficient electronic devices.

Researchers at the Indian Institute of Technology Guwahati have developed a trailblazing Machine Learning framework known as 'LEAP' to advance Electronic Design Automation (EDA). This novel approach is set to enhance the design efficiency of Integrated Circuits (ICs), which are pivotal to the $600 billion semiconductor industry.

According to a press release from IIT Guwahati, the creation of ICs is heavily reliant on EDA software that converts high-level designs into a Graphic Design System (GDS) manufacturing format. Traditional methods, which focus on heuristic techniques for problem-solving, often fall short of achieving optimal results.

To overcome these limitations, Prof Chandan Karfa, Dr Sukanta Bhattacharjee, and BTech students from IIT Guwahati, along with Dr Animesh Basak Chowdhury of New York University, have crafted the LEAP framework. This technology significantly cuts down the number of configurations the mapping tool must evaluate by over 50%, streamlining the technology mapping process within EDA.

Highlighting the framework's impact, Prof Karfa stated, 'Our framework not only accelerates the mapping process but also enhances circuit performance by achieving a 2% reduction in clock period without increasing the circuit area.' Extensive testing on 21 designs demonstrated a 50% improvement in runtime and a significant reduction in the configurations reviewed.

The LEAP framework, which aligns with global performance standards while utilizing fewer configurations, has demonstrated profound implications for the semiconductor industry. This research underscores the potential for developing faster, more efficient, and energy-saving electronic devices, thereby driving technological innovation.

The promising findings have been published in the ACM/IEEE International Conference on Computer-Aided Design (ICCAD 2024), signaling a significant leap forward for Electronic Design Automation.

(With inputs from agencies.)