Session

DescriptionAs data-intensive applications push conventional computing to its limits, in-memory computing emerges as a transformative paradigm to overcome critical memory bottlenecks. This session explores cutting-edge advances in compute-in-memory architectures, showcasing their potential to revolutionize AI acceleration, large-scale similarity search, and Boolean satisfiability solving. We delve into innovative solutions including energy-efficient vector search in NAND flash and DRAM (using UPMEM), adaptive mixed-signal in-memory SAT solvers, and novel CAM- and LUT-based accelerators. The session also highlights state-of-the-art simulation frameworks and stochastic computing techniques that significantly enhance programmability and efficiency.