A New Way of Designing Computer Chips

Computer chips are the foundation of technology. From smart phones to computers to GPUs, one of the driving forces behind our technical lives can be drilled down to the power of a computer chip. Engineers are constantly seeking ways to make computer chips faster, with more processing cores, and lower power consumption to keep up with consumer demands.

Wei Li, a Ph.D. candidate in the electrical and computer engineering department at Carnegie Mellon University, is developing a novel graph and large language model (LLM) based approach to optimize the chip design process to make artificial intelligence (AI) applications more efficient.

Every action on a digital device, from launching an app to playing a game, is translated into machine instructions for the graphics processing unit (GPU) or processor to execute. These instructions consist of sequences of binary digits (1s and 0s). A simple task might require thousands of instructions, while a complex operation, like gaming, can execute billions of them in seconds.

Fundamental hardware components that process these instructions include logic gates. A logic gate is a microscopic circuit that performs a basic logical function. For example, a common type known as an ‘AND’ gate will only output a '1' when all of its inputs are '1'; otherwise, its output is '0'.

“The millions of logic gates on a single chip are interconnected in a vast network that closely resembles a complex graph structure,” explains Li.

This representation is key to modern innovation.

“By viewing the chip's design as a graph, we can apply artificial intelligence and LLMs to analyze and improve it,” explains Li. “My career goal is to develop and revolutionize the next generation of design and test flow methods, ultimately driving better performing computer chips.”

Li’s novel graph-supported LLM for EDA tasks is just one of many innovative projects that he is working on. Due to his forward thinking and passion for inventing new processes for technology design, his co-advisors, Shawn Blanton, the Joseph F. and Nancy Keithley Professor of Electrical and Computer Engineering, and José M. F. Moura, the Philip L. and Marsha Dowd University Professor of Electrical and Computer Engineering, refer to him as “New Way.”

“I continue to be in awe of Wei Li’s research ideas and execution,” says Blanton. “He is constantly finding new ways to solve challenging problems. That mindset is crucial in today’s ever changing technology world. We are fortunate to have Wei in the Carnegie Mellon Engineering family.”

Li tributes his research ethos to his numerous internships at prestigious companies, as well as the opportunity to study under multiple distinguished professors at different universities.

“You can say my new ways are really coming from those experiences,” says Li.

The recipient of many prestigious fellowships and awards, Li was a highly sought-after Ph.D. candidate. He ultimately enrolled at Carnegie Mellon University because of the varied domains offered by the Department of Electrical and Computer Engineering.

“My advisors grant me the opportunity to learn from both ends of the computer chip spectrum,” says Li. “Professor Blanton focuses on the physical aspect of computer chips, while Professor Moura is an expert in signal graph processing. It’s a unique offering to be advised by skilled professors in both domains, which positively impacts my research findings.”

Drawing from both hardware and algorithmic perspectives, Li is developing chip designs that could help deliver more powerful and reliable electronics for everyone.

“Since joining CMU and Professor Blanton's and my research group, Wei has had an immediate and significant impact,” says Moura. “Most recently, he has advanced the field of Electronic Design Automation by developing a novel framework that integrates a graph-based approach with LLMs. His innovative work has captured the attention and support from several leading electronic manufacturers.”

When asked about his future, Li looks forward to continuing his cutting edge research as a university professor.

“In research, you are really doing something that may have never been done before,” Li says. “I really enjoy the sense that you are doing something brand new that will contribute to society.”