Paper Abstract and Keywords |
Presentation |
2021-08-18 13:00
[Invited Talk]
A 12nm autonomous driving processor running 60.4 TOPS and 13.8 TOPS/W CNNs with task-separated ASIL D control Katsushige Matsubara, Lieske Hanno (Renesas Electronics), Motoki Kimura (Renesas Electronics Europe), Atsushi Nakamura, Manabu Koike, Kazuaki Terashima, Shun Morikawa, Yoshihiko Hotta, Takahiro Irita, Seiji Mochizuki, Hiroyuki Hamasaki, Tatsuya Kamei (Renesas Electronics) SDM2021-39 ICD2021-10 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
Next-generation driver assistance systems and automated driving systems require both high performances to realize enormous computation and air-coolable power consumption to reduce the weight and cost of automotive electronic control units. In addition, the ISO 26262 functional safety standard requires that the highest safety level, ASIL D, be achieved. Redundancy is one of the measures to meet ASIL D but solving the significant increase in area and power consumption is a challenge. In this paper, we describe a convolutional neural network (CNN) accelerator that achieves both performance and power efficiency on the test chip, a hardware supported software lockstep mechanism applied to the CNN accelerator. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Automated driving processor / CNN hardware / ISO 26262 / functional safety / ASIL D / Freedom from Interference / / |
Reference Info. |
IEICE Tech. Rep., vol. 121, no. 139, ICD2021-10, pp. 48-53, Aug. 2021. |
Paper # |
ICD2021-10 |
Date of Issue |
2021-08-10 (SDM, ICD) |
ISSN |
Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
SDM2021-39 ICD2021-10 |