ISSCC 2024 Session 30 Overview: Domain-Specific Computing and Digital Accelerators

Advanced digital circuits techniques continue to drive innovations in domain-specific computing and digital accelerators.This session highlights six papers to improve the state-of-the-art energy efficiency and system performance for domainspecific and emerging applications.The first paper demonstrates a low-retention power, distributed nonvolatile-memory-based accelerator targeting micro-surveillance.The next two papers showcase the usage of compute-in-memory based processing units for implementing energy efficient spiking neural network and Boolean satisfiability problem solvers, respectively, followed by two designs of annealing processors for solving combinatorial optimization problems efficiently.The last paper demonstrates a RISC-V vector co-processor with integrated computein-memory vector register file to enable new architectural opportunities for data reuse.