<i>Nebula</i>: A Scalable and Flexible Accelerator for DNN Multi-Branch Blocks on Embedded Systems
Deep neural networks (DNNs) are widely used in many artificial intelligence applications; many specialized DNN-inference accelerators have been proposed. However, existing DNN accelerators rely heavily on certain types of DNN operations (such as Conv, FC, and ReLU, etc.), which are either less used...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-02-01
|
| Series: | Electronics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2079-9292/11/4/505 |
| _version_ | 1797480949260222464 |
|---|---|
| author | Dawei Yang Xinlei Li Lizhe Qi Wenqiang Zhang Zhe Jiang |
| author_facet | Dawei Yang Xinlei Li Lizhe Qi Wenqiang Zhang Zhe Jiang |
| author_sort | Dawei Yang |
| collection | DOAJ |
| description | Deep neural networks (DNNs) are widely used in many artificial intelligence applications; many specialized DNN-inference accelerators have been proposed. However, existing DNN accelerators rely heavily on certain types of DNN operations (such as Conv, FC, and ReLU, etc.), which are either less used or likely to become out of date in future, posing challenges of flexibility and compatibility to existing work. This paper designs a flexible DNN accelerator from a more generic perspective rather than speeding up certain types of DNN operations. Our proposed <i>Nebula</i> exploits the width property of DNNs and gains a significant improvement in system throughput and energy efficiency over multi-branch architectures. <i>Nebula</i> is a first-of-its-kind framework for multi-branch DNNs. |
| first_indexed | 2024-03-09T22:07:31Z |
| format | Article |
| id | doaj.art-7cdf87ccee0747ab894db3ff0a2c21c1 |
| institution | Directory Open Access Journal |
| issn | 2079-9292 |
| language | English |
| last_indexed | 2024-03-09T22:07:31Z |
| publishDate | 2022-02-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Electronics |
| spelling | doaj.art-7cdf87ccee0747ab894db3ff0a2c21c12023-11-23T19:38:19ZengMDPI AGElectronics2079-92922022-02-0111450510.3390/electronics11040505<i>Nebula</i>: A Scalable and Flexible Accelerator for DNN Multi-Branch Blocks on Embedded SystemsDawei Yang0Xinlei Li1Lizhe Qi2Wenqiang Zhang3Zhe Jiang4Academy for Engineering and Technology, Fudan University, Shanghai 200433, ChinaSchool of Statistics and Information, Shanghai University of International Business and Economics, Shanghai 201620, ChinaAcademy for Engineering and Technology, Fudan University, Shanghai 200433, ChinaAcademy for Engineering and Technology, Fudan University, Shanghai 200433, ChinaComputer Science, University of Cambridge, Cambridge CB3 0FD, UKDeep neural networks (DNNs) are widely used in many artificial intelligence applications; many specialized DNN-inference accelerators have been proposed. However, existing DNN accelerators rely heavily on certain types of DNN operations (such as Conv, FC, and ReLU, etc.), which are either less used or likely to become out of date in future, posing challenges of flexibility and compatibility to existing work. This paper designs a flexible DNN accelerator from a more generic perspective rather than speeding up certain types of DNN operations. Our proposed <i>Nebula</i> exploits the width property of DNNs and gains a significant improvement in system throughput and energy efficiency over multi-branch architectures. <i>Nebula</i> is a first-of-its-kind framework for multi-branch DNNs.https://www.mdpi.com/2079-9292/11/4/505DNN acceleratorsmulti-branch networkenergy-efficient accelerators |
| spellingShingle | Dawei Yang Xinlei Li Lizhe Qi Wenqiang Zhang Zhe Jiang <i>Nebula</i>: A Scalable and Flexible Accelerator for DNN Multi-Branch Blocks on Embedded Systems Electronics DNN accelerators multi-branch network energy-efficient accelerators |
| title | <i>Nebula</i>: A Scalable and Flexible Accelerator for DNN Multi-Branch Blocks on Embedded Systems |
| title_full | <i>Nebula</i>: A Scalable and Flexible Accelerator for DNN Multi-Branch Blocks on Embedded Systems |
| title_fullStr | <i>Nebula</i>: A Scalable and Flexible Accelerator for DNN Multi-Branch Blocks on Embedded Systems |
| title_full_unstemmed | <i>Nebula</i>: A Scalable and Flexible Accelerator for DNN Multi-Branch Blocks on Embedded Systems |
| title_short | <i>Nebula</i>: A Scalable and Flexible Accelerator for DNN Multi-Branch Blocks on Embedded Systems |
| title_sort | i nebula i a scalable and flexible accelerator for dnn multi branch blocks on embedded systems |
| topic | DNN accelerators multi-branch network energy-efficient accelerators |
| url | https://www.mdpi.com/2079-9292/11/4/505 |
| work_keys_str_mv | AT daweiyang inebulaiascalableandflexibleacceleratorfordnnmultibranchblocksonembeddedsystems AT xinleili inebulaiascalableandflexibleacceleratorfordnnmultibranchblocksonembeddedsystems AT lizheqi inebulaiascalableandflexibleacceleratorfordnnmultibranchblocksonembeddedsystems AT wenqiangzhang inebulaiascalableandflexibleacceleratorfordnnmultibranchblocksonembeddedsystems AT zhejiang inebulaiascalableandflexibleacceleratorfordnnmultibranchblocksonembeddedsystems |