Monolithically integrated asynchronous optical recurrent accelerator

Abstract Computing with light is widely recognized as a promising paradigm for overcoming the energy and latency limitations of electronic computing. However, the energy consumption and latency in current optical computing hardware predominantly arise in the electrical domain rather than the optical...

Full description

Saved in:
Bibliographic Details
Main Authors: Bo Wu, Haojun Zhou, Junwei Cheng, Wenkai Zhang, Shiji Zhang, Chaoran Huang, Dongmei Huang, Hailong Zhou, Jianji Dong, Xinliang Zhang
Format: Article
Language:English
Published: SpringerOpen 2025-05-01
Series:eLight
Subjects:
Online Access:https://doi.org/10.1186/s43593-025-00084-y
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849728419980378112
author Bo Wu
Haojun Zhou
Junwei Cheng
Wenkai Zhang
Shiji Zhang
Chaoran Huang
Dongmei Huang
Hailong Zhou
Jianji Dong
Xinliang Zhang
author_facet Bo Wu
Haojun Zhou
Junwei Cheng
Wenkai Zhang
Shiji Zhang
Chaoran Huang
Dongmei Huang
Hailong Zhou
Jianji Dong
Xinliang Zhang
author_sort Bo Wu
collection DOAJ
description Abstract Computing with light is widely recognized as a promising paradigm for overcoming the energy and latency limitations of electronic computing. However, the energy consumption and latency in current optical computing hardware predominantly arise in the electrical domain rather than the optical domain, primarily due to frequent signal conversions between optical (analog) and electrical (digital) formats. Furthermore, as the operating frequency of optical computing surpasses the GHz range, the synchronization of parallel electrical signals and the management of optical delays become increasingly critical. These challenges exacerbate energy consumption and latency, particularly in recurrent optical operations. To address these limitations, we propose a novel asynchronous computing paradigm for on-chip optical recurrent accelerators based on wavelength encoding, effectively mitigating synchronization challenges. By leveraging the intrinsic causality of wavelength relay, our approach eliminates the need for rigorous temporal alignment. To demonstrate the flexibility and efficacy of this asynchronous paradigm, we present two advanced recurrent models—an optical hidden Markov model and an optical recurrent neural network—monolithically integrated for the first time. These models incorporate hundreds of linear and nonlinear computing units densely packed into a compact footprint of just 10 mm2. Experimental evaluations on various benchmark tasks underscore the superior energy efficiency and low latency of the proposed asynchronous optical accelerators. This innovation enables the efficient processing of large-scale parallel signals and positions optical processors as a pivotal technology for applications such as autonomous driving and intelligent robotics.
format Article
id doaj-art-c107b8fc8b7d4f8d8b681e28f2b94cfc
institution DOAJ
issn 2097-1710
2662-8643
language English
publishDate 2025-05-01
publisher SpringerOpen
record_format Article
series eLight
spelling doaj-art-c107b8fc8b7d4f8d8b681e28f2b94cfc2025-08-20T03:09:34ZengSpringerOpeneLight2097-17102662-86432025-05-015111410.1186/s43593-025-00084-yMonolithically integrated asynchronous optical recurrent acceleratorBo Wu0Haojun Zhou1Junwei Cheng2Wenkai Zhang3Shiji Zhang4Chaoran Huang5Dongmei Huang6Hailong Zhou7Jianji Dong8Xinliang Zhang9Wuhan National Laboratory for Optoelectronics, School of Optical and Electronic Information, Huazhong University of Science and TechnologyWuhan National Laboratory for Optoelectronics, School of Optical and Electronic Information, Huazhong University of Science and TechnologyWuhan National Laboratory for Optoelectronics, School of Optical and Electronic Information, Huazhong University of Science and TechnologyWuhan National Laboratory for Optoelectronics, School of Optical and Electronic Information, Huazhong University of Science and TechnologyWuhan National Laboratory for Optoelectronics, School of Optical and Electronic Information, Huazhong University of Science and TechnologyDepartment of Electronic Engineering, The Chinese University of Hong KongPhotonics Research Institute, Department of Electrical and Electronic Engineering, The Hong Kong Polytechnic UniversityWuhan National Laboratory for Optoelectronics, School of Optical and Electronic Information, Huazhong University of Science and TechnologyWuhan National Laboratory for Optoelectronics, School of Optical and Electronic Information, Huazhong University of Science and TechnologyWuhan National Laboratory for Optoelectronics, School of Optical and Electronic Information, Huazhong University of Science and TechnologyAbstract Computing with light is widely recognized as a promising paradigm for overcoming the energy and latency limitations of electronic computing. However, the energy consumption and latency in current optical computing hardware predominantly arise in the electrical domain rather than the optical domain, primarily due to frequent signal conversions between optical (analog) and electrical (digital) formats. Furthermore, as the operating frequency of optical computing surpasses the GHz range, the synchronization of parallel electrical signals and the management of optical delays become increasingly critical. These challenges exacerbate energy consumption and latency, particularly in recurrent optical operations. To address these limitations, we propose a novel asynchronous computing paradigm for on-chip optical recurrent accelerators based on wavelength encoding, effectively mitigating synchronization challenges. By leveraging the intrinsic causality of wavelength relay, our approach eliminates the need for rigorous temporal alignment. To demonstrate the flexibility and efficacy of this asynchronous paradigm, we present two advanced recurrent models—an optical hidden Markov model and an optical recurrent neural network—monolithically integrated for the first time. These models incorporate hundreds of linear and nonlinear computing units densely packed into a compact footprint of just 10 mm2. Experimental evaluations on various benchmark tasks underscore the superior energy efficiency and low latency of the proposed asynchronous optical accelerators. This innovation enables the efficient processing of large-scale parallel signals and positions optical processors as a pivotal technology for applications such as autonomous driving and intelligent robotics.https://doi.org/10.1186/s43593-025-00084-yAsynchronous operationOptical recurrent acceleratorOptical hidden Markov modelOptical recurrent neural network
spellingShingle Bo Wu
Haojun Zhou
Junwei Cheng
Wenkai Zhang
Shiji Zhang
Chaoran Huang
Dongmei Huang
Hailong Zhou
Jianji Dong
Xinliang Zhang
Monolithically integrated asynchronous optical recurrent accelerator
eLight
Asynchronous operation
Optical recurrent accelerator
Optical hidden Markov model
Optical recurrent neural network
title Monolithically integrated asynchronous optical recurrent accelerator
title_full Monolithically integrated asynchronous optical recurrent accelerator
title_fullStr Monolithically integrated asynchronous optical recurrent accelerator
title_full_unstemmed Monolithically integrated asynchronous optical recurrent accelerator
title_short Monolithically integrated asynchronous optical recurrent accelerator
title_sort monolithically integrated asynchronous optical recurrent accelerator
topic Asynchronous operation
Optical recurrent accelerator
Optical hidden Markov model
Optical recurrent neural network
url https://doi.org/10.1186/s43593-025-00084-y
work_keys_str_mv AT bowu monolithicallyintegratedasynchronousopticalrecurrentaccelerator
AT haojunzhou monolithicallyintegratedasynchronousopticalrecurrentaccelerator
AT junweicheng monolithicallyintegratedasynchronousopticalrecurrentaccelerator
AT wenkaizhang monolithicallyintegratedasynchronousopticalrecurrentaccelerator
AT shijizhang monolithicallyintegratedasynchronousopticalrecurrentaccelerator
AT chaoranhuang monolithicallyintegratedasynchronousopticalrecurrentaccelerator
AT dongmeihuang monolithicallyintegratedasynchronousopticalrecurrentaccelerator
AT hailongzhou monolithicallyintegratedasynchronousopticalrecurrentaccelerator
AT jianjidong monolithicallyintegratedasynchronousopticalrecurrentaccelerator
AT xinliangzhang monolithicallyintegratedasynchronousopticalrecurrentaccelerator