OpenVX高效能并行可重构运算通路的设计与实现

doi:10.19678/j.issn.1000-3428.0060764

摘要/Abstract

摘要： 针对专用硬件在处理图形图像时无法同时兼顾灵活性、可扩展性和时效性的问题，设计一种支持OpenVX 1.3标准的专用处理器。通过对OpenVX 1.3标准中的核函数进行数据通路映射，分析实现函数高效处理所需的运算单元数目，确定适用于该标准的数据通路运算器的结构。通过编写指令对数据通路进行重构，适应OpenVX标准的演进和扩展。应用65 nm CMOS工艺库对整体电路进行综合验证，实现的OpenVX可重构数据通路运算器面积为21 076.21 μm²、功耗为778.63 mW、系统主频为500 MHz、吞吐量为1.86 GB/s。实验结果表明，该数据通路运算器具有较强的可编程性和可扩展性，能够有效满足实时和高速的通用图像处理要求。

关键词: 图像处理, 计算机视觉, OpenVX标准, 并行处理, 可重构

Abstract: When used for graphics and image processing, application-specific hardware usually fail to balance flexibility, scalability and speed.To address the problem, an application-specific instruction set processor supporting the OpenVX 1.3 standard is designed.Through the data path mapping of the kernel function in OpenVX 1.3 and the analysis of the number of operation units needed to realize the efficient processing of the function, the suitable computing path structure for the standard is determined.The data path is reconstructed by writing instructions to adapt to the evolution and expansion of openvx standard.The overall circuit is verified by using 65nm CMOS process Library.The area of the OpenVX reconfigurable computing pathway is 21 076.2 μm², the power consumption is 778.63 mW, the main frequency of the system is 500 MHz, and the system throughput is 1.86 GB/s.The experimental results show that the data path operator has strong programmability and scalability, and can effectively meet the requirements of real-time and high-speed general image processing.

Key words: image processing, computer vision, OpenVX standard, parallel processing, reconfigurability

中图分类号:

TP311

王宇, 李涛, 邢立冬, 冯臻夫. OpenVX高效能并行可重构运算通路的设计与实现[J]. 计算机工程, 2021, 47(12): 236-248.

WANG Yu, LI Tao, XING Lidong, FENG Zhenfu. Design and Implementation of OpenVX Highly Efficient Parallel Reconfigurable Computing Pathway[J]. Computer Engineering, 2021, 47(12): 236-248.

http://www.ecice06.com/CN/Y2021/V47/I12/236

图/表 23

20211214182245

20211214182249

20211214182252

20211214182255

20211214182258

20211214182302

20211214182305

20211214182308

20211214182312

20211214182315

20211214182318

20211214182322

20211214182325

20211214182328

20211214182333

20211214182337

20211214182341

20211214182345

20211214182350

20211214182354

20211214182359

20211214182404

20211214182409

参考文献

[1] 李涛, 孙建, 王鹏博.基于PAAG的OpenVX核心库函数并行化实现[J].西安邮电大学学报, 2015, 20(2):7-10. LI T, SUN J, WANG P B.Parallel implementation of kernels of OpenVX based on PAAG[J].Joural of Xi'an University of Posts and Telecommunications, 2015, 20(2):7-10.(in Chinese)
[2] RADHAKRISHNA G.The khronos OpenVX working group[EB/OL].[2021-01-01].https://www.khronos.org/registry/OpenVX/specs/1.3/html/OpenVX_Specification_1_3.html.
[3] BRILL F, ERUKHIMOV V, GIDUTHURI R, et al.Deploying an OpenVX graph to a target platform[M].OpenVX Programming Guide, 2020:65-84.
[4] 延酉玫, 李涛, 王鹏博.OpenVX与三维渲染在多态GPU上的并行实现[J].计算机应用, 2015, 35(1):53-57. YAN Y M, LI T, WANG P B, et al.Parallel implementation of OpenVX and 3D rendering on polymorphic graphics processing unit[J].Joural of Computer Applications, 2015, 35(1):53-57.(in Chinese)
[5] 黄灿.基于OpenVX的图像预处理算法的并行化研究[J].现代计算机, 2020(34):36-39. HUANG C.Parallel research of image pre-processing algorithm based on OpenVX[J].Modern Computer, 2020(34):36-39.(in Chinese)
[6] TAGLIAVINI G, HAOUGOU G, MARONGIU A, et al.ADRENALINE:an OpenVX environment to optimize embedded vision applications on many-core accelerators[C]//Proceedings of the 9th International Symposium on Embedded Multicore/Many-core Systems-on-Chip.Washington D.C., USA:IEEE Press, 2015:289-296.
[7] SAJJAD T, PAYMAN B, ELI B, et al.AFFIX:automatic acceleration framework for FPGA implementation of OpenVX vision algorithms[C]//Proceedings of 2019 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays.New York, USA:ACM Press, 2019:252-261.
[8] ABEYSINGHE M, VILLARREAL J, WEAVER L, et al.OpenVX graph optimization for visual processor units[C]//Proceedings of the 30th International Conference on Application-Specific Systems, Architectures and Processors.Washington D.C., USA:IEEE Press, 2019:123-130.
[9] TAHERI S, HEO J, BEHNAM P, et al.Acceleration framework for FPGA implementation of OpenVX graph pipelines[C]//Proceedings of the 26th Annual International Symposium on Field-Programmable Custom Computing Machines.Washington D.C., USA:IEEE Press, 2018:227-227.
[10] 李岑, 贺光辉.用于实时目标检测的FPGA神经网络加速器设计[J].微电子学与计算机, 2020, 37(7):6-11. LI C, HE G H.Design of FPGA-based neural network accelerator for real-time objective detection[J].Microelectronics and Computer, 2020, 37(7):6-11.(in Chinese)
[11] 邵杰, 万书芹, 任凤霞.基于ASIC的并行流水线级联半带滤波器设计[J].固体电子学研究与进展, 2020, 40(1):60-65. SHAO J, WAN S Q, REN F X.An ASIC based parallel pipelined cascade half band filter design[J].Research and Progress of SSE, 2020, 40(1):60-65.(in Chinese)
[12] 吴皓月, 邓军勇, 山蕊, 等.可重构阵列处理器Harris算法并行化实现[J].微电子学与计算机, 2019, 36(4):67-71. WU H Y, DENG J Y, SHAN R, et al.Reconfigurable array processor harris algorithm is implemented in parallel[J].Microeletronics and Computer, 2019, 36(4):67-71.(in Chinese)
[13] 邓文齐, 郑启龙, 盛鑫, 等.分簇架构处理器上卷积并行计算算法的研究[J].小型微型计算机系统, 2018, 39(3):520-524. DENG W Q, ZHENG Q L, SHENG X, et al.Study of convolution parallel algorithm on multi-cluster processor[J].Journal of Chinese Computer Systems, 2018, 39(3):520-524.(in Chinese)
[14] 章子凯, 武继刚, 姜文超, 等.容错处理器阵列的多逻辑列并行重构算法[J].计算机工程与科学, 2018, 40(1):24-33. ZHANG Z K, WU J G, JIANG W C, et al.A parallel multiple logical columns reconfiguration algorithm on fault-tolerant processor arrays[J].Computer Engineering and Science, 2014, 40(1):24-33.(in Chinese)
[15] 李涛, 杨婷, 易学渊, 等.萤火虫2:一种多态并行机的硬件体系结构[J].计算机工程与科学, 2014, 36(2):191-200. LI T, YANG T, YI X Y, et al.Firefly 2:a hardware architecture for polymorphic parallel computers[J].Computer Engineering and Science, 2014, 36(2):191-200.(in Chinese)
[16] 高向强, 冯春阳, 闫鑫, 等.一种面向64位DSP处理器的可重构ALU研究及设计[J].微电子学与计算机, 2015, 32(10):1-6. GAO X Q, FENG C Y, YAN X, et al.Research and design of a reconfigurable ALU for 64 bit digital signal processor[J].Microelectronics and Computer, 2015, 32(10):1-6.(in Chinese)
[17] 张嘉琛, 蒋剑飞, 毛志刚.基于功能复用的高性能ALU设计[J].信息技术, 2010, 34(3):58-60, 63. ZHANG J C, JIANG J F, MAO Z G.Design of an efficient ALU based on reused logic structure[J].Information Technology, 2010, 34(3):58-60, 63.(in Chinese)
[18] 刘容, 赵洪深, 李晓今.基于改进型选择进位加法器的32位浮点乘法器设计[J].现代电子技术, 2013, 36(16):133-136. LIU R, ZHAO H S, LI X J.Design of 32-bit floating-point multiplier based on improved carry-select adder[J].Modern Electronics Technique, 2013, 36(16):133-136.(in Chinese)
[19] 宋子豪, 李涛, 杜晓鸽, 等.光栅化中多格式除法器的设计与实现[J].电子世界, 2020(5):114-115. SONG Z H, LI T, DU X G.Design and implementation of multi format divider in rasterization[J].Electronics World, 2020(5):114-115.(in Chinese)
[20] 何军, 黄永勤, 朱英.一种高性能四倍精度浮点乘加器的设计与实现[J].计算机工程, 2014, 40(2):294-299. HE J, HUANG Y Q, ZHU Y.Design and implementation of a high performance quadruple precision floating-point multiplier accumulator[J].Computer Engineering, 2014, 40(2):294-299.(in Chinese)
[21] 车文博, 刘衡竹, 田甜.M-DSP中高性能浮点乘加器的设计与实现[J].计算机应用, 2016, 36(8):2213-2218. CHE W B, LIU H Z, TIAN T.Design and implementation of high performance floating-point multiply acculate for M-DSP[J].Journal of Computer Applications, 2016, 36(8):2213-2218.(in Chinese)

选择文件类型/文献管理软件名称

选择包含的内容