Reetuparna Das receives ACM SIGARCH Maurice Wilkes Award

Reetuparna Das, associate professor of computer science and engineering at U-M, has been selected to receive the 2024 Maurice Wilkes Award by the ACM Special Interest Group on Computer Architecture (SIGARCH). The award recognizes Das’s outstanding achievements, research, and leadership in the field of computer architecture.

Given usually to one individual annually who has been producing research for no more than 20 years, the Maurice Wilkes Award is a high honor for mid-career researchers in computer architecture. The award’s namesake, Maurice Wilkes, was a renowned computer scientist credited with inventing microprogramming and designing one of the world’s first stored-program computers.

Das is an established expert in computer architecture, with a record of cutting-edge research. Broadly, her work focuses on the interplay between software systems and device technologies, including in-memory architectures, custom computing for precision health applications, heterogeneous processor architecture for mobile systems, and low-power scalable interconnects.

Prof. Das received the Maurice Wilkes Award for her contributions to in-memory computing. In today’s memory structures, over two-thirds of a processor is dedicated to cache, and all of the main memory is devoted to temporary storage. None of these memory elements can compute, but could they? Das’s research addresses this fundamental question regarding the role of memory, and proposes to impose a dual responsibility on them: to store and compute data.

She and her collaborators have proposed techniques that repurpose cache memory arrays into over a million compute units with an increase of only ~3% in CPU size. This style of in-memory computing repurposes existing caches into massive vector compute units whose parallelism is several orders of magnitude higher than a contemporary GPU. Additionally, this memory structure saves energy spent shuffling data between storage and compute units—a significant concern in big data applications. In-memory computing is a significant departure from processing-in-memory (PIM) technologies that move computing power closer to memory (near-memory) but do not compute data in place.

Das pioneered In-SRAM computing with her collaborators Scott Mahlke and Satish Narayanasamy of CSE, as well as David Blaauw and Dennis Sylvester of Electrical and Computer Engineering at U-M. She started this line of work with her paper “Compute caches” (HPCA 2017), widely considered a groundbreaking idea in the field. Das’s subsequent work, “Neural cache: Bit-serial in-cache acceleration of deep neural networks” (ISCA 2018), was the first to demonstrate in-memory arithmetic operation and showed its efficiency for convolutional neural networks (CNNs). Subsequently, she and her collaborators successfully demonstrated the circuit feasibility of a transposed bit-serial design with a chip prototype, as reported in their paper “14.2 A compute SRAM with bit-serial integer/floating-point operations for programmable in-memory vector acceleration” (ISSCC 2019).

Das’s work on neural cache was selected for IEEE Micro Top Picks and featured in a retrospective of the top research in the field from over the past 25 years published by the International Symposium on Computer Architecture (ISCA) in 2023. Prof. Das’ group continues to advance PIM technology and publish papers in premier venues. She also wrote a book titled “In-/Near-Memory Computing,” published as a Synthesis Lecture on Computer Architecture by Morgan & Claypool.

Today, in-memory computing is a prominent research area across computer architecture and related fields, with a significant amount of research being published on this topic and multiple startups commercializing In-SRAM computing for edge AI. Das has collaborated with cache expert and Intel Fellow Ravi Iyer on this work, and neural cache remains a promising idea on Intel’s bucket list.

Das was recognized with Intel’s Outstanding Researcher Award for this line of work. She has also received a Sloan Foundation Faculty Fellowship, CRA-W Borg Early Career Award, and an NSF CAREER Award. In addition, she was recently elected chair of the ACM Special Interest Group on Microarchitecture (SIGMICRO), and she previously served as vice-chair of ACM SIGMICRO, chair of the MICRO test-of-time award committee, and program co-chair of the 2019 MICRO conference.

Currently, at U-M, Das is serving as a founding member of the MAVERIC collaborative, which is leading the university’s participation in the CHIPS and Science Act. She is also working on a $15-million project with Los Alamos National Laboratory focused on advancing exascale computing.

Before joining the faculty at CSE, Das was a researcher at Intel Labs and with the Center for Future Architectures Research. She also co-founded a precision medicine startup called Sequal Inc. She completed her PhD in Computer Science and Engineering at Pennsylvania State University, University Park.