NoC加速器中的高效DNN动态切片与智能映射算法

齐芸; 欧阳一鸣

doi:10.11959/j.issn.1000-0801.2025179

您当前的位置：

首页 >

文章列表页 >

NoC加速器中的高效DNN动态切片与智能映射算法

研究与开发 | 更新时间：2025-11-06

- NoC加速器中的高效DNN动态切片与智能映射算法
- Efficient DNN dynamic slicing and intelligent mapping algorithm in NoC accelerator
- 电信科学 2025年41卷第10期页码：151-160
- 作者机构：
  
  1.安徽交通职业技术学院，安徽合肥 230051
  2.合肥工业大学计算机与信息学院，安徽合肥 230051
- 作者简介：
  
  [ "齐芸（1984- ），女，安徽交通职业技术学院讲师，主要研究方向为片上网络。" ]
  [ "欧阳一鸣（1963- ），男，博士，合肥工业大学教授、博士生导师，主要研究方向为基于片上网络的人工智能应用。" ]
- 基金信息：
  
  国家自然科学基金资助项目(62374049);安徽高校自然科学研究项目(2024AH050281;2024AH040051;2024AH050284)
- DOI：10.11959/j.issn.1000-0801.2025179
  中图分类号： TP183
- 收稿：2025-03-17，
  
  修回：2025-04-30，
  
  录用：2025-06-04，
  
  纸质出版：2025-10-20
- 稿件说明：
移动端阅览
齐芸,欧阳一鸣.NoC加速器中的高效DNN动态切片与智能映射算法[J].电信科学,2025,41(10):151-160.

QI Yun,OUYANG Yiming.Efficient DNN dynamic slicing and intelligent mapping algorithm in NoC accelerator[J].Telecommunications Science,2025,41(10):151-160.
齐芸,欧阳一鸣.NoC加速器中的高效DNN动态切片与智能映射算法[J].电信科学,2025,41(10):151-160. DOI： 10.11959/j.issn.1000-0801.2025179.

QI Yun,OUYANG Yiming.Efficient DNN dynamic slicing and intelligent mapping algorithm in NoC accelerator[J].Telecommunications Science,2025,41(10):151-160. DOI： 10.11959/j.issn.1000-0801.2025179.

摘要

针对深度神经网络（deep neural network，DNN）模型在传统切片与映射方法中存在的资源调度和数据传输瓶颈问题，提出了一种基于片上网络（network on chip，NoC）加速器的高效DNN动态切片与智能映射优化算法。该算法通过动态切片技术灵活划分DNN模型的计算任务，并结合智能映射策略优化NoC架构中的任务分配与数据流管理。实验结果表明，与传统方法相比，该算法在计算吞吐量、NoC传输时延、外部内存访问次数和计算能效等方面均显著提升，尤其在复杂模型上表现突出。

Abstract

To address the bottlenecks of resource scheduling and data transmission in traditional slicing and mapping methods for deep neural networks (DNN)

an efficient dynamic slicing and intelligent mapping optimization algorithm was proposed based on a network on chip (NoC) accelerator. The algorithm was designed to flexibly divide DNN computing tasks through dynamic slicing and optimize task and data flow management in the NoC architecture. Experimental results show that the proposed algorithm significantly outperforms traditional methods in computing throughput

NoC transmission delay

external memory accesses

and energy efficiency

especially for complex models.

关键词

Keywords

references

CHEN Y H , YANG T J , EMER J , et al . Eyeriss v2: a flexible accelerator for emerging deep neural networks on mobile devices [J ] . IEEE Journal on Emerging and Selected Topics in Circuits and Systems , 2019 , 9 ( 2 ): 292 - 308 .

LIU Z , WU H , YU X , et al . Neu-NoC: neural-inspired network-on-chip architecture for energy-efficient AI computing [J ] . IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems , 2021 , 40 ( 5 ), 887 - 900 .

KIM J , KANG S , PARK J . Adaptive task mapping for energy-efficient NoC-based deep learning accelerators [J ] . ACM Transactions on Design Automation of Electronic Systems , 2020 , 25 ( 4 ): 1 - 19 .

LEE J , KIM C , KANG S , et al . UNPU: an energy-efficient deep neural network accelerator with fully variable weight bit precision [J ] . IEEE Journal of Solid-State Circuits , 2018 , 54 ( 1 ): 173 - 185 .

HOJABR S , NAJAFI M , FALLAH F . A congestion-aware router for power-efficient network-on-chip architectures [J ] . IEEE Transactions on Very Large Scale Integration (VLSI) Systems , 2020 , 28 ( 12 ): 2698 - 2708 .

JIANG J . WANG P, XIE Y. Memory access pattern analysis in NoC-based DNN accelerators using booksim simulator [J ] . IEEE Transactions on Parallel and Distributed Systems , 2021 , 32 ( 3 ): 678 - 692 .

YU X , TANG X , XU C , WANG Y . Towards efficient DNN inference on resource-constrained edge devices: a network-on-chip perspective [J ] . IEEE Internet of Things Journal , 2020 , 7 ( 9 ): 8653 - 8666 .

LI H , CHEN Y , WANG Z . Energy-efficient mapping of deep neural networks on NoC-based accelerators [J ] . Journal of Systems Architecture , 2020 , 108 : 101741 .

ZHANG Z , ZHOU H , CHEN W . Power-aware task scheduling for deep learning accelerators on chip-multiprocessors [J ] . Journal of Parallel and Distributed Computing , 2021 , 151 : 42 - 54 .

PATEL R , SHARMA P , GUPTA S . Reinforcement learning-based dynamic mapping for NoC-based deep learning accelerators [J ] . Neurocomputing , 2020 , 387 : 91 - 103 .

ZHAO Y , LIU X , ZHANG W . A survey of memory optimization techniques for deep learning accelerators [J ] . IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems , 2021 , 40 ( 7 ): 1325 - 1338 .

TAN M X , QUOC Le . Efficientnet: rethinking model scaling for convolutional neural networks [C ] // Proceedings of the 36th International Conference on Machine Learning (ICML) , Long Beach : PMLR , 2019 , 6105 - 6114 .

WANG L , ZHANG Y , LI X . A survey of network-on-chip architectures for deep learning accelerators [J ] . IEEE Transactions on Computers , 2020 , 69 ( 8 ): 1234 - 1248 .

GUO K , ZENG S , CHEN T . Dynamic resource allocation for deep learning tasks in NoC-based systems [J ] . IEEE Transactions on Parallel and Distributed Systems , 2021 , 32 ( 6 ): 1345 - 1358 .

XU J , WANG H , CHEN L . Congestion-aware task mapping for NoC-based deep learning accelerators [J ] . IEEE Transactions on Computers , 2020 , 69 ( 12 ): 1876 - 1889 .

ZHANG X , LI Y , WANG J . Reinforcement learning for dynamic resource allocation in NoC-based deep learning systems [J ] . IEEE Transactions on Neural Networks and Learning Systems , 2021 , 32 ( 10 ): 4567 - 4579 .

浏览量

159

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

暂无数据