The University of Michigan – Shanghai Jiao Tong University Joint Institute (UM-SJTU JI, JI hereafter) research team led by Assistant Professor Xinfei Guo has won 2022 Best Paper Award at the Semiconductor Technology Forum with the article “A Multi-Functional SRAM-Based Compute-in-Memory Macro on the Edge” first-authored by undergraduate student Runxi Wang. The forum co-hosted by Chinese tech giant Huawei Corporation and Fudan University was held in Shanghai last month.

 

The research project of the team was also selected to participate in the 10th SJTU Chun-Tsung Program. In the annual simulated international conference sponsored by the program, Runxi Wang was invited as the only student speaker sharing the status quo of the project. She was offered the Prominent Performance Award during the conference.

 

Compute-in-memory (CIM hereafter) is one of the trending paradigms to overcome the “Memory Wall” bottleneck in the conventional Von Neumann computer architectures. It offers potential benefits of reducing the energy consumption in the next-generation edge AI chips and other portable devices. Since early 2022, Runxi Wang started working on the research project that focused on developing SRAM-based compute-in-memory architectures under the guidance of Professor Xinfei Guo. Different from other existing CIM work, her research concentrates more on extending the computing capability and introducing the reconfigurability into the existing CIM architectures. To embrace the increasing diversity and complexity of future edge devices, the goal of this research is to develop cross-layer approaches that enable the low-cost reconfigurability at circuit, architecture and system levels. Preliminary results out of this work have been accepted by multiple international conferences or forums such as ACM Student Research Competition (co-located with International Conference on Computer Aided Design) and TinyML Asia Forum, where Wang has presented orally. Wang plans to pursue a Ph.D. degree at JI upon receiving her bachelor’s degree in the summer of 2023 and conduct further research in the computer architecture related fields.

 

Best Paper Award Certificate of the Semiconductor Technology Forum

Illustration of SRAM-based Compute-in-Memory (CIM) Paradigm

 

“JI is a unique institute where it nurtures students with both hard and soft skills. When I started my freshman year, I was so nervous when I spoke English in public. But now I can handle English presentations in front of a large crowd very confidently. I got poor grades in the beginning, but now I am among top 10%. I used to be so confused about my future career plans, but now I find my research interests and become more clear about my next steps. JI has played a significant role in offering unique learning environment and numerous research opportunities for me, ” said Wang. During her undergraduate study, Wang worked as a teaching assistant for seven times for various courses and is a Chief Assistant Mentor of the JI Center of Learning and Teaching (CLT) in charge of training new teaching assistants. She is also an active participant of JI international exchange programs. She joined the winter program at the Ritsumeikan University in Japan in 2020 and will go to the Cornell University in the United States in January 2023 for an exchange program of one semester.

 

Speaking about the future, Wang said: “It was truly an amazing experience when I finally found my keen research interests in computer hardware, originated from taking the relevant courses. A taste of the research in my undergraduate study has shaped my future research directions. I believe it will be challenging yet precious to explore the unknowns as part of my research journey. But I am fully ready for it and looking forward to advancing the research fields I am in. In the meanwhile, I will keep my passion for the humanity subjects to enrich my life.”