AERO: A 1.28 MOP/s/LUT Reconfigurable Inference Processor for Recurrent Neural Networks in a Resource-Limited FPGA

AERO: A 1.28 MOP/s/LUT Reconfigurable Inference Processor for Recurrent Neural Networks in a Resource-Limited FPGA

This study presents a resource-efficient reconfigurable inference processor for recurrent neural networks (RNN), named AERO. AERO is programmable to perform inference on RNN models of various types. This was designed based on the instruction-set architecture specializing in processing primitive vect...

Full description

Bibliographic Details
Main Authors: Jinwon Kim, Jiho Kim, Tae-Hwan Kim
Format: Article
Language:English
Published: MDPI AG 2021-05-01
Series:Electronics
Subjects:
accelerator architectures
field programmable gate arrays
microarchitecture
neural network hardware
recurrent neural networks
Online Access:https://www.mdpi.com/2079-9292/10/11/1249

Internet

https://www.mdpi.com/2079-9292/10/11/1249

Similar Items