On-the-Fly Lowering Engine: Offloading Data Layout Conversion for Convolutional Neural Networks

Many deep learning frameworks utilize GEneral Matrix Multiplication (GEMM)-based convolution to accelerate CNN execution. GEMM-based convolution provides faster convolution yet requires a data conversion process called lowering (i.e., im2col), which incurs significant memory overhead and diminishes...

Полное описание

Библиографические подробности

Главные авторы:	Mingu Kang, Sangmin Hyun, Tae Hee Han, Jungrae Kim, Seokin Hong
Формат:	Artigo
Язык:	Inglês
Опубликовано:	IEEE 2022-01-01
Серии:	IEEE Access
Предметы:	Convolutional neural network GEMM CPU
Online-ссылка:	https://ieeexplore.ieee.org/document/9833499/
Метки:	Добавить метку Нет меток, Требуется 1-ая метка записи!