Researchers at the Indian Institute of Technology Guwahati (IITG) have achieved a significant breakthrough in Electronic Design Automation (EDA) by developing an innovative machine learning (ML) framework called ‘LEAP’.

This advanced solution optimises the design process of Integrated Circuits (ICs), a vital element in the $600 billion semiconductor industry that underpins modern electronic devices.

The creation of ICs is heavily dependent on EDA software, which converts high-level designs into a manufacturing format known as a Graphic Design System (GDS).

However, designing ICs presents complex challenges that are often difficult to resolve. Traditional methods frequently rely on heuristic techniques—efficient problem-solving strategies that provide satisfactory solutions, though not necessarily the optimal ones. While these methods help balance design quality with runtime, they often result in suboptimal outcomes.

To overcome these challenges, Prof. Chandan Karfa, Associate Professor, and Dr. Sukanta Bhattacharjee, Assistant Professor from the Department of Computer Science and Engineering at IIT Guwahati, along with their BTech students Chandrabhushan Reddy Chigarapally and Harshwardhan Nitin Bhakkad, have utilised machine learning to enhance efficiency in IC design.

Collaborating with Dr. Animesh Basak Chowdhury from New York University, USA, they developed the LEAP framework, which streamlines the technology mapping process within EDA. Instead of evaluating thousands of potential configurations, LEAP intelligently identifies and prioritises the most promising options, reducing the number of configurations the mapping tool must consider by over 50%.

Prof Chandan Karfa said, “Our framework not only speeds up the mapping process but also improves the performance of the circuits. We have reduced the runtime of the EDA tool by 50% and achieved a 2% reduction in clock period without increasing the area required for the circuits, making our solution a significant advancement in electronic design automation.”

LEAP estimates delays for various configurations and selects only the top ten options for each node in the design, compared to traditional methods that typically evaluate around 250 configurations. This targeted approach simplifies the workflow and boosts overall efficiency.

In extensive testing across 21 different designs, the LEAP framework achieved a 50% improvement in runtime and reduced the number of configurations checked by over 51%. Compared to exhaustive mapping methods, LEAP delivers similar performance results while using 63% fewer configurations, significantly enhancing the runtime of the open-source ABC EDA tool.

This research has significant implications for the semiconductor industry, which is crucial for the development of electronic devices such as smartphones and computers.

By optimising the IC design process with the LEAP framework, researchers can cut design time and enhance performance. This results in faster, more efficient electronic devices with reduced energy consumption, benefiting consumers and fostering innovation across different technology sectors.

The promising outcomes of this research have been published in the ACM/IEEE International Conference on Computer-Aided Design (ICCAD 2024).

Image source: X