Specification of generic Memristor-based Crossbar Architectures for In-Memory Computing

Seiler B, Fey D (2025)

Publication Type: Conference contribution

Publication year: 2025

Event location: Dresden, Sachsen DE

Journal Issue: 2025

DOI: 10.1109/MOCAST65744.2025.11083925

Abstract

Processing data where it resides, as envisioned by in-memory computing (IMC) concepts, is expected to be the next step after multi-core and heterogeneous architectures consisting of CPU and accelerators to counteract the energy wall. The use of memristors offers low read energy and non-volatility of the data when accessing the memory. This helps energy- and computing performance hungry embedded functions for artificial intelligence and machine learning. In addition, memristive devices enable the integration of storage and processing functions in close proximity to significantly reduce the energy required to transport data. To simplify the design and simulation of IMC architectures, mixed-signal or generally mixed-abstraction simulations would be ideal, which simultaneously combine digital architecture elements with analogue components, e.g. memristive matrix-vector accelerators. This requires new simulation environments that enable fast simulation times and a flexible design process. To move forward in this direction, we present our SystemC-based solution for the design and simulation of memristive crossbar architectures and demonstrate and evaluate its usefulness for selected neuromorphic architectures.

Authors with CRIS profile

Dietmar Fey Lehrstuhl für Informatik 3 (Rechnerarchitektur)

How to cite

APA:

Seiler, B., & Fey, D. (2025). Specification of generic Memristor-based Crossbar Architectures for In-Memory Computing. In Proceedings of the IEEE 14th International Conference on Modern Circuits and Systems Technologies (MOCAST). Dresden, Sachsen, DE.

MLA:

Seiler, Benjamin, and Dietmar Fey. "Specification of generic Memristor-based Crossbar Architectures for In-Memory Computing." Proceedings of the IEEE 14th International Conference on Modern Circuits and Systems Technologies (MOCAST), Dresden, Sachsen 2025.

BibTeX: Download