Article
A Heterogeneous RISC-V Processor for Efficient DNN
Application in Smart Sensing System
Haifeng Zhang
1
, Xiaoti Wu
2,3,4
, Yuyu Du
3,5
, Hongqing Guo
3,6
, Chuxi Li
3,4,5
, Yidong Yuan
1
,
Meng Zhang
3,4,5,
* and Shengbing Zhang
3,4,5
Citation: Zhang, H.; Wu, X.; Du, Y.;
Guo, H.; Li, C.; Yuan, Y.; Zhang, M.;
Zhang, S. A Heterogeneous RISC-V
Processor for Efficient DNN
Application in Smart Sensing System.
Sensors 2021, 21, 6491. https://
doi.org/10.3390/ s21196491
Academic Editors: Sotiris Kotsiantis,
Panagiotis E. Pintelas and
Ioannis E. Livieris
Received: 30 August 2021
Accepted: 25 September 2021
Published: 28 September 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1
National & Local Joint Engineering Research Center for Reliability Technology of Energy Internet Intelligent
Terminal Core Chip, Beijing Smart-Chip Microelectronics Technology Co., Ltd., Beijing 100192, China;
zhanghaifeng@sgitg.sgcc.com.cn (H.Z.); yuanyidong@sgitg.sgcc.com.cn (Y.Y.)
2
School of Cybersecurity, Northwestern Polytechnical University, Xi’an 710072, China;
xiw26@mail.nwpu.edu.cn
3
Engineering and Research Center of Embedded Systems Integration (Ministry of Education),
Xi’an 710129, China; 2019262266@mail.nwpu.edu.cn (Y.D.); guohongqing@mail.nwpu.edu.cn (H.G.);
lichuxi@mail.nwpu.edu.cn (C.L.); zhangsb@nwpu.edu.cn (S.Z.)
4
National Engineering Laboratory for Integrated Aero-Space-Ground-Ocean Big Data Application Technology,
Xi’an 710129, China
5
School of Computer Science, Northwestern Polytechnical University, Xi’an 710129, China
6
School of Software, Northwestern Polytechnical University, Xi’an 710129, China
* Correspondence: zhangm@nwpu.edu.cn
Abstract:
Extracting features from sensing data on edge devices is a challenging application for
which deep neural networks (DNN) have shown promising results. Unfortunately, the general
micro-controller-class processors which are widely used in sensing system fail to achieve real-time
inference. Accelerating the compute-intensive DNN inference is, therefore, of utmost importance.
As the physical limitation of sensing devices, the design of processor needs to meet the balanced
performance metrics, including low power consumption, low latency, and flexible configuration.
In this paper, we proposed a lightweight pipeline integrated deep learning architecture, which is
compatible with open-source RISC-V instructions. The dataflow of DNN is organized by the very
long instruction word (VLIW) pipeline. It combines with the proposed special intelligent enhanced
instructions and the single instruction multiple data (SIMD) parallel processing unit. Experimental
results show that total power consumption is about 411 mw and the power efficiency is about
320.7 GOPS/W.
Keywords: sensing system; dnn; intelligent computing architecture; RISC-V; VLIW; SIMD
1. Introduction
Traditional Internet of Things (IoT) devices are usually responsible for data mea-
surement, data collection, and pre-processing tasks. Due to the limitation of bandwidth,
the huge amount of data generated by the edge devices cannot be transmitted to the cloud
for further AI intelligent computing. Extracting features from sensing data by DNN in the
sensing system is challenging as deploying intelligent applications requires the trade-off
between real-time and high efficiency in the resource-limited edge devices. At this stage,
the widely-used micro-controller-class processors in sensing system, such as MCS51 and
STM32, accomplish given tasks without an operating system and with limited memory and
low processing capacity. Because of the poor performance of the micro control unit (MCU),
deploying neural networks directly on micro-controller-class processors faces many diffi-
culties. Notably, intelligent applications impose strict requirements on: (1) high computing
performance, (2) low power consumption, and (3) flexible configuration [
1
]. Therefore, it
is necessary to design advanced processors equipped to the sensing system to satisfy the
demands of deploying DNN with balanced performance metrics. Therefore, it is necessary
Sensors 2021, 21, 6491. https://doi.org/10.3390/s21196491 https://www.mdpi.com/journal/sensors
