Contribution to a conference proceedings

FZJ-2024-01857

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Zeroth and higher-order logic with content addressable memories

Pedretti, G. ; Böhm, F. ; Hizzani, M.FZJ* ; Bhattacharya, T. ; Bruel, P. ; Moon, J. ; Serebryakov, S. ; Strukov, D. ; Strachan, J. P.FZJ* ; Ignowski, J. ; Van Vaerenbergh, T. ; Beausoleil, R. ; Mohseni, M.

2023
IEEE

This record in other databases:

Please use a persistent id in citations: doi:10.1109/IEDM45741.2023.10413853

Abstract: Content Addressable Memories (CAMs) are attracting interest as in-memory computational primitives, thanks to the massively parallel search operation. Multiple flavors of CAMs have been realized with nanoscale memory technology, pushing their performance towards low power and latency. In this work, we demonstrate how to use CAMs for asserting and solving zeroth order and higher-order logic, in the form of Boolean satisfiability (SAT) and satisfiability modulo theories (SMT), respectively. We demonstrate a ~6.5× lower area and ~4× lower energy per search compared with state-of-the-art in-memory optimization problem solvers, such as Hopfield Neural Network (HNN), with up to 175× faster time-to-solution for problems with 150 variables.

---

Contributing Institute(s):

1. Neuromorphic Compute Nodes (PGI-14)

Research Program(s):

1. 5234 - Emerging NC Architectures (POF4-523) (POF4-523)

Appears in the scientific report 2024

---