An Energy-Efficient Near-Memory Computing Architecture for CNN Inference at Cache Level
Nouripayam M, Prieto A, Kishorelal VK, Rodrigues J (2021)
Publication Type: Conference contribution
Publication year: 2021
Publisher: Institute of Electrical and Electronics Engineers Inc.
Conference Proceedings Title: 2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings
Event location: Dubai, ARE
ISBN: 9781728182810
DOI: 10.1109/ICECS53924.2021.9665530
Abstract
A non-von Neumann Near-Memory Computing architecture, optimized for CNN inference in edge computing, is integrated in the cache memory sub-system of a microcontroller unit. The NMC co-processor is evaluated using an 8-bit fixed-point quantized CNN model, and achieves an accuracy of 98% on the MNIST dataset. A full inference of the CNN model executed on the NMC processor, demonstrates an improvement of more than 34× in performance, and 28× in energy-efficiency, compared to the baseline scenario of a conventional single-core processor. The design achieves a performance of 1.39 GOPS (at 200 MHz) and an energy-efficiency of 49 GOPS/W, with negligible area overhead of less than 1%.
Involved external institutions
Sweden (SE)How to cite
APA:
Nouripayam, M., Prieto, A., Kishorelal, V.K., & Rodrigues, J. (2021). An Energy-Efficient Near-Memory Computing Architecture for CNN Inference at Cache Level. In 2021 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021 - Proceedings. Dubai, ARE: Institute of Electrical and Electronics Engineers Inc..
MLA:
Nouripayam, Masoud, et al. "An Energy-Efficient Near-Memory Computing Architecture for CNN Inference at Cache Level." Proceedings of the 28th IEEE International Conference on Electronics, Circuits, and Systems, ICECS 2021, Dubai, ARE Institute of Electrical and Electronics Engineers Inc., 2021.
BibTeX: Download