Tutorial: Advances and Prospects for In-Memory Computing

As applications become more data centric, the cost of accessing data from memory dominates energy and throughput. An emerging paradigm to address this is in-memory computing. Early approaches show the potential to reduce energy-delay product (EDP) by 2-to-3 orders of magnitude. However, architectures for enabling computation within the constrained structure of memory raise new challenges and opportunities. This tutorial will map out advances in these areas and examine recent demonstrations.