计算机科学 ›› 2016, Vol. 43 ›› Issue (8): 318-322.doi: 10.11896/j.issn.1002-137X.2016.08.065
• 图形图像与模式识别 • 上一篇
张静,熊承义,高志荣
ZHANG Jing, XIONG Cheng-yi and GAO Zhi-rong
摘要： 针对大尺度压缩感知重构算法实时性应用的需要,探讨了基于图形处理器(GPU)的正交匹配追踪算法(OMP)的加速方法及实现。为降低中央处理器与GPU之间传输的高延迟,将整个OMP算法的迭代过程转移到GPU上并行执行。其中,在GPU端根据全局存储器的访问特点,改进CUDA程序使存储访问满足合并访问条件,降低访问延迟。同时,根据流多处理器(SM)的资源条件,增加SM中共享存储器的分配,通过改进线程访问算法来降低bank conflict,提高访存速度。在NVIDIA Tesla K20Xm GPU和Intel(R) E5-2650 CPU上进行了测试,结果表明,算法中耗时长的投影模块、更新权值模块分别可获得32和46倍的加速比,算法整体可获得34倍的加速比。
| [1] Donoho D L.Compressed sensing [J].IEEE Transactions on Information Theory (S0018-9448),2006,52(4):1289-1306 [2] Shao Wen-ze,Wei Zhi-hui.Advances and perspectives on compressed sensing theory[J].Journal of Image and Graphics,2012,7(1):1-12(in Chinese) 邵文泽,韦志辉.压缩感知基本理论:回顾与展望[J].中国图象图形学报,2012,17(1):1-12 [3] Jiao Li-cheng,Yang Shu-yuan,Liu Fang,et al.Development and Prospect of Compressive Sensing[J].Acta Electronica Sinica,2011,9(7):1651-1662(in Chinese) 焦李成,杨淑媛,刘芳,等.压缩感知回顾与展望[J].电子学报,2011,39(7):1651-1662 [4] Dai Qiong-hai,Fu Chang-jun,Ji Xiang-yang.Research on Compressed Sensing[J].Chinese Journal of Computers,2011,4(3):425-434(in Chinese) 戴琼海,付长军,季向阳.压缩感知研究[J].计算机学报,2011,34(3):425-434 [5] Mueller K,Yagel R.On the Use of Graphics Hardware to Acce-lerate Algebraic Reconstruction Methods[C]∥Proceedings of SPIE Medical Imaging Conference 1999.San Diego,America,1999 [6] Xiong Z,Chi W Q,Lu K,et al.GPU Acceleration of Saliency Detection Algorithm[C]∥Proceedings of the 11th International Symposium on Distributed Computing and Applications to Business,Engineering & Science.IEEE,Guangxi,2012:48-51 [7] Guo Rui-ran,Song Jian-xin.Research and Implementation of Pa-rallel Signal Reconstruction about Image Compressed Sensing Based on GPU[J].Video Engineering,2014,8(11):15-19(in Chinese) 郭睿冉,宋建新.图象压缩感知的基于GPU的并行重构算法研究[J].电视技术,2014,38(11):15-19 [8] Li Guo-yan,Hou Xiang-dan,Gu Jun-hua,et al.A GPU-BasedParallel Design and Implementation of Sparse MRI Reconstruction Algorithm[J].Computer Applications and Software,2013,0(9):163-166(in Chinese) 李国燕,侯向丹,顾军华,等.稀疏磁共振图像重建算法的 GPU 并行设计与实现[J].计算机应用与软件,2013,30(9):163-166 [9] Kulkarni A,Mohsenin T.Accelerating Compressive Sensing Reconstruction OMP Algorithm with CPU,GPU,FPGA and Domain Specific Many-Core[C]∥IEEE International Symposium on Circuits and Systems(ISCAS 2015).Lisbon,Portugal,2015:970-973 [10] Fowler J E,Mun S,Tramel E W.Block-Based Compressed Sen-sing of Images and Video [J].Foundations Trends in Signal Processing,2012,4:297-416 [11] Lee H,Oh H,Lee S,et al.Visually Weighted Compressive Sen-sing:Measurement and Reconstruction [J].IEEE Trans.on Image Processing,2013,22(4):1444-1455 [12] Farber R.CUDA Application Design and Development [M].Burlington:Morgan Kaufmann,2011 [13] Wang Ze-huan,Wang Peng.Introduction to CPU Parallel Programming Technology[J].E-science Technology & Application,2013,4(1):81-87(in Chinese) 王泽寰,王鹏.GPU并行计算编程技术介绍[J].科研信息化技术与应用,2013,4(1):81-87 [14] 张舒,褚艳利.GPU高性能运算之CUDA[M].北京:中国水利水电出版社,2009 [15] Deng Chun-yuan,Wei Yi-min.Some New Results of the Sherman-Morrison-Woodbury Formula[C]∥Proceeding of The Sixth International Conference of Matrices and Operators.Chengdu,China,2011,2:220-223 [16] Hager W W.Updating the Inverse of a Matrix [J].SIAM REVIEW,1989,31(2):221-239 |
| No related articles found! |
|
||
首页