Abstract:
Processing-in-Memory (PIM) promises to break free from the data movement bottleneck by performing computation where the data resides. While early DRAM-based PIMs handled only simple operations, recent research shows that even complex and irregular workloads can be efficiently executed inside unmodified DRAM arrays.

This talk explores two such breakthroughs. NTT-PIM performs the Number-Theoretic Transform (NTT) entirely in memory through clever row-centric mapping and pipelined buffering, achieving up to 17× speedup over prior work. Building on that idea, SPIMA extends PIM to sparse matrix multiplication (SpMM) using a novel dataflow that maximizes bank-level parallelism and data reuse for highly sparse data.

Together, these works showcase how PIM is evolving---from accelerating structured transforms to enabling scalable sparse computation---toward a future where memory is not just for storage, but also for compute.

Bio:
Jongeun Lee received his B.Sc. and M.Sc. degrees in Electrical Engineering, and his Ph.D. in Electrical Engineering and Computer Science, all from Seoul National University, Korea. He is currently a Professor in the Department of Electrical Engineering at the Ulsan National Institute of Science and Technology (UNIST), Ulsan, Korea, which he joined in 2009. Before joining UNIST, he was a postdoctoral research associate at Arizona State University, Tempe, Arizona, USA, and a researcher at Samsung Electronics, Korea.

He has published over 100 peer-reviewed journal and conference papers and has served on the technical program and organizing committees of several international conferences and workshops in the areas of computer-aided design, embedded systems, and reconfigurable computing.

His research interests include hardware/software co-design, reconfigurable architectures, deep learning acceleration, and computer-aided design for emerging technologies.