面向FT-M7002平台点积算法的优化实现

计算机工程与科学 ›› 2022, Vol. 44 ›› Issue (11): 1909-1917.

面向FT-M7002平台点积算法的优化实现

郭盼盼1,2,陈梦雪3,梁祖达1,2,马晓畅3，许邦建4

(1.郑州大学计算机与人工智能学院，河南郑州 450066；2.国家超级计算郑州中心(郑州大学)，河南郑州 450001；
3.湖南大学电气与信息工程学院，湖南长沙 410082；4.湖南大学信息科学与工程学院，湖南长沙 410082)

收稿日期:2022-02-07 修回日期:2022-04-01 接受日期:2022-11-25 出版日期:2022-11-25 发布日期:2022-11-25

Optimization of dot product algorithms on FT-M7002

GUO Pan-pan1,2,CHEN Meng-xue3,LIANG Zu-da1,2,MA Xiao-chang3,XU Bang-jian4

(1.School of Computer and Artificial Intelligence,Zhengzhou University,Zhengzhou 450066；
2.National Supercomputing Center in Zhengzhou(Zhengzhou University),Zhengzhou 450001;
3.School of Electrical and Information Engineering,Hunan University,Changsha 410082;
4.School of Information Science and Engineering,Hunan University,Changsha 410082,China)

Received:2022-02-07 Revised:2022-04-01 Accepted:2022-11-25 Online:2022-11-25 Published:2022-11-25

摘要/Abstract

摘要： 基于国产的FT-M7002平台高性能DSP，针对不同类型的点积算法进行了优化实现,完善了该处理器平台数学库的技术链,充分发挥了FT-M7002内核体系结构优势，对点积算法实现了SIMD向量并行化、DMA双通道传输和SVR传输等优化。该研究充分挖掘了程序的向量并行性，有效地提升了数据传输的速度，提高了程序性能。实验结果表明，输入不同规模大小的数组，不同类型的点积算法在FT-M7002平台上优化后和优化前的平均性能比为12.416 6~45.233 8。相较于TI官网的dsplib库中不同类型的点积函数在TMS320C6678处理器上运行的性能，FT-M7002平台优化后的性能与TI平台的平均性能比为1.371 6~4.519 6。实验结果表明了该DSP平台相对于TI主流平台的计算性能优势。

关键词: FT-M7002, DSP, 点积算法, 向量, DMA双通道传输, SVR传输

Abstract: On the high-performance DSP of domestic FT-M7002 platform, different types of dot product algorithms are optimized and implemented. The technical chain of the mathematical library of the processor platform is consummated. Taking full advantage of FT-M7002 kernel architecture, SIMD vector parallelization, DMA dual channel transmission, SVR transmission and other optimization methods for dot product algorithm are realized. The research fully excavates the vector parallelism of the program, effectively improving the speed of data transmission and improving the performance of the program. The experimental results show that the average performance ratio of different types of dot product algorithms after and before optimization on FT platform is 12.416 6~45.233 8. Compared with the performance of different types of dot product functions in dsplib library on TI official website on TMS320C6678 processor, the average performance ratio between FT platform and TI platform is 1.371 6 ~ 4.519 6. The research results show that the DSP platform has obvious computational performance advantages over TI mainstream platform.

Key words: FT-M7002, digital signal processor(DSP), dot product algorithm, vector, DMA dual channel transmission, SVR transmission

郭盼盼, 陈梦雪, 梁祖达, 马晓畅, 许邦建. 面向FT-M7002平台点积算法的优化实现[J]. 计算机工程与科学, 2022, 44(11): 1909-1917.

GUO Pan-pan, CHEN Meng-xue, LIANG Zu-da, MA Xiao-chang, XU Bang-jian. Optimization of dot product algorithms on FT-M7002[J]. Computer Engineering & Science, 2022, 44(11): 1909-1917.

参考文献［22］

［1］	Yuan Guo-xing,Zhang Yun-quan,Yuan Liang.State of the art analysis of China HPC 2021 ［J］.Computer Engineering & Science,2021,43 (12):2091-2097.(in Chinese)
［2］	Huang Chun,Jiang Hao,Gu Tong-xiang,et al.Implementation and optimization of high-precision summation and dot product algorithms on Phythium processor ［J］.Computer Engineering & Science,2021,43(1):1-8.(in Chinese)
［3］	Xie Qing-chun,Zhang Yun-quan,Wang Ke,et al.Research of the SIMD and vector math library ［J］.Computer Science,2011,38(7):298-301.(in Chinese)
［4］	Parhami B.SIMD machines:Do they have a significant future? ［J］.SIGARCH Computer Architecture News,1995,23(4):19-22.
［5］	Xun Z Y,Ding H F,He Z.Practical time optimization of DSP program for high-frequency applications［J］.Journal of Physics:Conference Series,2021,2095:012005.
［6］	Han X T,Mou J,Xiong L,et al.Coexistence of infinite attractors in a fractional-order chaotic system with two nonlinear functions and its DSP implementation［J］.Integration,2021,81:43-55.
［7］	Li L,Zhang Y,Guo Q,et al.Real-time object tracking method based on multi-core DSP［J］.Journal of Physics:Confe- rence Series,2021,1873:012029.
［8］	Zhou Pei,Zhou Wei-chao,Wang Kai-kai.Research on performance of TMS320C6678 multicore DSP parallel memory access［J］.Microcomputer & Its Applications,2014,33(13):20-24.(in Chinese)
［9］	Zhao Yi-pei.Development and application of DSP［J］.Elec-tronic Technology & Software Engineering,2018(6):92.(in Chinese)
［10］	Blanchard P,Higham N J,Mary T.A class of fast and accurate summation algorithms［J］.SIAM Journal on Scientific Computing,2020,42(3):A1541-A1557.
［11］	Gai Min.Research on design and application of dot product protocol based on secure multi-party computation［D］.Jinan:Shandong University,2021.(in Chinese)
［12］	Carbó-Dorca R.Generalized scalar products in Minkowski metric spaces［J］.Journal of Mathematical Chemistry,2021,59(4):1029-1045.
［13］	Xiao Yun-hui,Liu Qun,Li Cheng-en.Limit laws for the number of triangles in the sparse random dot product graphs［J］.Journal of Minnan Normal University(Natural Science),2021,34(1):48-52.(in Chinese)
［14］	Gao Wei,Luo Yi-xuan,Li Jia-kun,et al.High-performance hardware implementation of elliptic curve cryptography point multiplication over GF(p)［J］.Transactions of Beijing Institute of Technology,2021,41(9):977-984.(in Chinese)
［15］	Sun Guang-hui.OpenCV transplantation and optimization based on FT-M7002［D］.Xi’an:Xidian University,2019.(in Chinese)
［16］	Chen Yun,Wang Meng-yuan,Chai Xiao-nan,et al.Optimization of Gaussian filtering algorithm on FT-M7002［J］.Computer Engineering & Science,2021,43(5):799-806.(in Chinese)
［17］	Guo Heng-liang,Chai Xiao-nan,Han Lin,et al.Realization and optimization of Canny edge detection in FT platform［J］.Computer Engineering,2021,47(7):37-43.(in Chinese)
［18］	Fan Ming-liang,Guo Zi-han,Chai Xiao-nan,et al.Optimized realization of Sobel edge detection algorithm for FT-M7002 ［J］.Computer Engineering,2022,48(6):193-199.(in Chinese)
［19］	Zhang Shuai. Design and implementation of a DMA host-controller to support multi-transmission mode［D］.Changsha:National University of Defense Technology,2014.(in Chinese)
［20］	Xu Ao-li.Design and implementation of Matrix2 configurable scalar data memory［D］.Changsha:National University of Defense Technology,2014.(in Chinese)
［21］	Department of Mathematics,Tongji University.Engineering mathematics:Linear algebra［M］.6th ed. Beijing:Higher Education Press,2014.(in Chinese)
［22］	Hu Yue-an.Design and verification of the DMA to support efficient transmission in 32-bit high-performance M-DSP［D］.Changsha:National University of Defense Technology,2015.(in Chinese)
	附中文参考文献：
［1］	袁国兴,张云泉,袁良.2021年中国高性能计算机发展现状分析［J］.计算机工程与科学,2021,43(12):2091-2097.
［2］	黄春,姜浩,谷同祥,等.面向飞腾处理器的高精度求和与点乘算法实现和优化［J］.计算机工程与科学,2021,43(1):1-8.
［3］	解庆春,张云泉,王可,等.SIMD技术与向量数学库研究［J］.计算机科学,2011,38(7):298-301.
［8］	周佩,周维超,王凯凯.TMS320C6678多核DSP并行访问存储器性能的研究［J］.微型机与应用,2014,33(13):20-24.
［9］	赵一霈.DSP的发展与应用［J］.电子技术与软件工程,2018(6):92.
［11］	盖敏.基于安全多方计算的点积协议设计与应用研究［D］.济南：山东大学,2021.
［13］	肖云辉,刘群,李城恩.稀疏随机点积图中三角形数量的极限定理［J］.闽南师范大学学报(自然科学版),2021,34(1):48-52.
［14］	高巍,骆宜萱,李佳琨,等.素数域椭圆曲线密码点乘的高性能硬件实现［J］.北京理工大学学报,2021,41(9):977-984.
［15］	孙广辉.基于FT-M7002的OpenCV移植与优化［D］.西安：西安电子科技大学,2019.
［16］	陈云,王梦园,柴晓楠,等.面向FT-M7002的高斯滤波算法优化实现［J］.计算机工程与科学,2021,43(5):799-806.
［17］	郭恒亮,柴晓楠,韩林,等.Canny边缘检测算法在飞腾平台上的实现与优化［J］.计算机工程,2021,47(7):37-43.
［18］	范明亮,郭子涵,柴晓楠,等.面向FT-M7002的Sobel边缘检测算法优化实现［J］.计算机工程,2022,48(6):193-199.
［19］	张帅.一种支持多种传输模式的DMA主机模块设计与实现［D］.长沙：国防科学技术大学,2014.
［20］	许奥利.Matrix2可配置标量数据存储器的设计及实现［D］.长沙：国防科学技术大学,2014.
［21］	同济大学数学系.工程数学:线性代数［M］.第六版.北京:高等教育出版社,2014.
［22］	胡月安.32位高性能M-DSP中支持高效数据传输的DMA设计与验证［D］.长沙：国防科学技术大学,2015.

[1]	田玉恒,马胜,鲁建壮,杨柳. 一种高效的DMA核间同步传输方法[J]. J4, 20160101, 38(01): 52-56.
[2]	胡庆孟, 王红斌, 王俊钟. 基于NPN融入词性注意力机制的中文事件探测[J]. 计算机工程与科学, 2023, 45(08): 1490-1497.
[3]	喻金平, 朱伟锋, 廖列法. 基于RoBERTa-wwm-BiLSTM-CRF的扶持政策文本实体识别研究[J]. 计算机工程与科学, 2023, 45(08): 1498-1507.
[4]	刘屹成, 刘晓燕, 严馨. 并行平衡级联支持向量机[J]. 计算机工程与科学, 2023, 45(07): 1170-1177.
[5]	李小玲, 方建滨, 马俊, 谭霜, 谭郁松. 基于监督学习的稀疏矩阵自动任务分配[J]. 计算机工程与科学, 2023, 45(05): 782-789.
[6]	排日旦·阿布都热依木, 吐尔地·托合提, 艾斯卡尔·艾木都拉, . 基于深度学习的实体关系抽取方法研究[J]. 计算机工程与科学, 2023, 45(05): 895-902.
[7]	董芃杉, 张晶, 金日泽. 基于双通道门控复合网络的中文产品评论情感分析[J]. 计算机工程与科学, 2023, 45(05): 911-919.
[8]	陈财森, 胡海荣, 程志炜, 房璐璐. 基于BA-RVM算法的发动机故障诊断技术研究[J]. 计算机工程与科学, 2023, 45(02): 332-337.
[9]	唐宇, 代琪, 杨梦园, 陈丽芳, . 改进麻雀搜索算法优化SVM的异常点检测[J]. 计算机工程与科学, 2023, 45(02): 346-354.
[10]	仲臣, 余学祥, 邰晓曼, 韩雨辰, 肖星星, 刘清华, . 萤火虫算法优化支持向量机室内定位研究[J]. 计算机工程与科学, 2022, 44(11): 1968-1975.
[11]	何涛, 施慧莉, 李大亮. 基于深度学习的SAR目标识别DSP设计[J]. 计算机工程与科学, 2022, 44(08): 1357-1363.
[12]	马汉达, 朱敏. 改进SVM不平衡数据分类的IGWOSMOTE方法[J]. 计算机工程与科学, 2022, 44(06): 1133-1140.
[13]	白皓, 甘新标, 杨文祥, 贾孟涵, 涂旭平, 张一鸣, 郭敏, 来乐, 张意, 朱春平. 面向大规模图计算的连通分量算法分析与优化[J]. 计算机工程与科学, 2022, 44(02): 191-198.
[14]	顾越, 赵银亮. 基于RISC-V向量指令的稀疏矩阵向量乘法实现与优化[J]. 计算机工程与科学, 2022, 44(01): 1-8.
[15]	范小康, 夏泽宇, 龙思凡, 杨灿群. 基于ARM SVE的光滑粒子流体动力学SIMD加速方法[J]. 计算机工程与科学, 2021, 43(06): 989-996.