基于GPU的数字下变频累积误差控制方法

doi:10.12305/j.issn.1001-506X.2023.04.04

摘要/Abstract

摘要：

针对数字下变频系统中高速数据处理实时性需求, 设计了基于中央处理器(central processing unit, CPU)+图形处理单元(graphic processing unit, GPU)的异构平台数字下变频处理算法, 并完成了相应并行程序设计, 进行了实际数据验证。鉴于GPU运算采用单精度浮点数格式进行数据处理, 进行长时程数据处理过程中, 会产生浮点数舍入误差累积, 导致结果的不准确不可靠, 设计了联合相位循环归零法和无误差变换补偿法, 实现了浮点数舍入误差的控制与补偿。数据测试结果表明, 所提方法可将相位误差控制在10^-11rad以内, 且不随时间累积, 计算结果准确可靠。

关键词: 图形处理单元, 舍入误差, 累积误差, 无误差变换

Abstract:

Aiming at the real-time demand of high-speed data processing in digital down-conversion system, a digital down-conversion processing algorithm based on central processing unit(CPU)+graphics processing unit(GPU) is designed for heterogeneous platform, and the corresponding parallel programming is completed, and the actual data is verified. Since GPU computing using single precision floating point number format on data processing, for long time history in the process of data processing, can produce floating-point rounding error accumulation, not accurate and reliable results. A combined phase cycle zero and error-free transformation compensation method is designed, and the floating-point number rounding error control and compensation is realized. Data test results show that the proposed method can control the error within 10^-11, and does not cumulative with time, and the calculation results are accurate and reliable.

Key words: graphics processing unit (GPU), round off error, cumulative error, error-free transformation

中图分类号:

TN911

李超, 焦义文, 傅诗媛, 高泽夫, 毛飞龙. 基于GPU的数字下变频累积误差控制方法[J]. 系统工程与电子技术, 2023, 45(4): 965-972.

Chao LI, Yiwen JIAO, Shiyuan FU, Zefu GAO, Feilong MAO. GPU-based digital down-conversion cumulative error control method[J]. Systems Engineering and Electronics, 2023, 45(4): 965-972.

图/表 13

表1

图1

图2

图3

图4

图5

图6

图7

图8

图9

图10

图11

表2

参考文献 36

1	SIKKA P , ASATI A R , SHEKHAR C . Power-and area-optimized high-level synthesis implementation of a digital down converter for software-defined radio applications[J]. Circuits, Systems, and Signal Processing, 2021, 40 (6): 2883- 2894. doi: 10.1007/s00034-020-01601-9
2	CARDILLO E , CANANZI R , VITA P , et al. Dual-conversion microwave down converter for nanosatellite electronic warfare systems[J]. Applied Sciences, 2022, 12 (3): 1524. doi: 10.3390/app12031524
3	NAGESH B K , REDDY V S . Digital down converter for 5G systems[J]. Perspectives in Communication, Embedded-Systems and Signal-Processing-PiCES, 2021, 7- 10.
4	DATTA D , MITRA P , DUTTA H S . FPGA implementation of high performance digital down converter for software defined radio[J]. Microsystem Technologies, 2022, 28 (2): 533- 542. doi: 10.1007/s00542-019-04579-w
5	LI Y , WANG M G , SUN J , et al. Photonic-assisted wideband frequency down-converter with leakage cancellation and image rejection for FMCW radar receiver[J]. IEEE Journal of Quantum Electronics, 2021, 58 (1): 8700107.
6	CHOI J , NA K I , SHIN Y , et al. Research on broadband signal processing techniques for the small millimeter wave tracking radar[J]. The Journal of the Institute of Internet, Broadcasting and Communication, 2021, 21 (6): 49- 55.
7	张凯, 陈龙, 秦奋, 等. 高倍抽取率的数字下变频设计[J]. 火控雷达技术, 2019, 48 (4): 33- 37.
	ZHANG K , CHEN L , QIN F , et al. Design of a digital down converter with high decimation rate[J]. Fire Control Radar Technology, 2019, 48 (4): 33- 37.
8	BOUTROS A , BETZ V . FPGA architecture: principles and progression[J]. IEEE Circuits and Systems Magazine, 2021, 21 (2): 4- 29. doi: 10.1109/MCAS.2021.3071607
9	MAZHER I J L , MANIKANDAN T . FPGA-based reconfigurable architectures for DSP computations[M]. Singapore: Springer, Advances in Smart System Technologies, 2021: 587- 594.
10	TIAN H W , GUO S , ZHAO P , et al. Design and implementation of a real-time multi-beam sonar system based on FPGA and DSP[J]. Sensors, 2021, 21 (4): 1425. doi: 10.3390/s21041425
11	KWIATKOWSKI P . Digital-to-time converter for test equipment implemented using FPGA DSP blocks[J]. Measurement, 2021, 177, 109267. doi: 10.1016/j.measurement.2021.109267
12	HALABI L M , ALSOFYANI I M , LEE K B . Hardware implementation for hybrid active NPC converters using FPGA-based dual pulse width modulation[J]. Journal of Power Electronics, 2021, 21 (11): 1669- 1679. doi: 10.1007/s43236-021-00305-w
13	LIU Y J, WANG J, LIU J. Accelerating the simulation of finite difference time domain (FDTD) with GPU[C]//Proc. of the IEEE International Joint EMC/SI/PI and EMC Europe Symposium, 2021: 707-711.
14	BERGER M , GIOTAKIS A I , PILLEI M , et al. Agreement between rhinomanometry and computed tomography-based computational fluid dynamics[J]. International Journal of Computer Assisted Radiology and Surgery, 2021, 16 (4): 629- 638. doi: 10.1007/s11548-021-02332-1
15	JEONG W K , FLETCHER P T , TAO R , et al. Interactive visualization of volumetric white matter connectivity in DT-MRI using a parallel-hardware Hamilton-Jacobi solver[J]. IEEE Trans.on Visualization and Computer Graphics, 2007, 13 (6): 1480- 1487. doi: 10.1109/TVCG.2007.70571
16	GADEKALLU T R , RAJPUT D S , REDDY M , et al. A novel PCA-whale optimization-based deep neural network model for classification of tomato plant diseases using GPU[J]. Journal of Real-Time Image Processing, 2021, 18 (4): 1383- 1396. doi: 10.1007/s11554-020-00987-8
17	ABBASI J S , BASHIR F , QURESHI K N , et al. Deep learning-based feature extraction and optimizing pattern matching for intrusion detection using finite state machine[J]. Computers & Electrical Engineering, 2021, 92, 107094.
18	JURCZUK K , CZAJKOWSKI M , KRETOWSKI M . Multi-GPU approach to global induction of classification trees for large-scale data mining[J]. Applied Intelligence, 2021, 51 (8): 5683- 5700. doi: 10.1007/s10489-020-01952-5
19	KUMAR V , SHARMA D K , MISHRA V K . Mille cheval: a GPU-based in-memory high-performance computing framework for accelerated processing of big-data streams[J]. The Journal of Supercomputing, 2021, 77 (7): 6936- 6960. doi: 10.1007/s11227-020-03508-3
20	SHI L . Application of big data language recognition technology and GPU parallel computing in English teaching visualization system[J]. International Journal of Speech Technology, 2022, 25 (3): 667- 677. doi: 10.1007/s10772-021-09904-1
21	ALLEN E, CHASE D, LUCHANGCO V, et al. Object-oriented units of measurement[C]//Proc. of the 19th Annual ACM SIGPLAN Conference on Object-oriented Programming, Systems, Languages, and Applications, 2004: 384-403.
22	OBERG J . Why the Mars probe went off course[J]. IEEE Spectrum, 1999, 36 (12): 34- 39. doi: 10.1109/6.809121
23	KAHAN W. How futile are mindless assessments of roundoff in floating-point computation?[EB/OL]. [2022-02-20]. http://http.cs.berkeley.edu/~kahan/Mindless.pdf.
24	VON N J , GOLDSTINE H H . Numerical inverting of matrices of high order[J]. Bulletin of the American Mathematical Society, 1947, 53 (11): 1021- 1099. doi: 10.1090/S0002-9904-1947-08909-6
25	TURING A M . Rounding-off errors in matrix processes[J]. The Quarterly Journal of Mechanics and Applied Mathematics, 1948, 1 (1): 287- 308. doi: 10.1093/qjmam/1.1.287
26	WILKINSON J H . Rounding errors in algebraic processes[M]. New York: Dover Publications, 1963.
27	HIGHAM N J . Accuracy and stability of numerical algorithm[M]. Philadelphia: Science for Industry and Applied Mathematics, 1996.
28	周毓麟, 袁国兴. 关于科学计算用数字电子计算机字长问题[J]. 计算机工程与科学, 2005, 27 (10): 1- 2.1-2, 16
	ZHOU Y L , YUAN G X . On the word length of digital electronic computers for scientific computations[J]. Computer Engineering & Science, 2005, 27 (10): 1- 2.1-2, 16
29	MULLER J M , BRISEBARRE N , DEDINECHIN F , et al. Handbook of floating-point arithmetic[M]. Basel, Switzerland: Birkhäuser, 2018.
30	姜浩. 高精度可靠浮点计算及舍入误差分析研究[D]. 长沙: 国防科学技术大学, 2013.
	JIANG H. Study on reliable computing and rounding[D]. Changsha: National University of Defense Technology, 2013.
31	杜配冰. 基于多部分浮点数表示格式的高精度算法研究[D]. 长沙: 国防科技大学, 2017.
	DU P B. Research on accurate algorithm based on multiple-component format of floating-point numbers[D]. Changsha: National University of Defense Technology, 2017.
32	OGITA T , RUMP S M , OISHI S . Accurate sum and dot pro-duct[J]. SIAM Journal on Scientific Computing, 2005, 26 (6): 1955- 1988. doi: 10.1137/030601818
33	RUMP S M , OGITA T , OISHI S . Accurate floating-point summation part Ⅰ: faithful rounding[J]. SIAM Journal on Scientific Computing, 2008, 31 (1): 189- 224. doi: 10.1137/050645671
34	RUMP S M , OGITA T , OISHI S . Accurate floating-point summation part Ⅱ: sign, K-fold faithful and rounding to nearest[J]. SIAM Journal on Scientific Computing, 2008, 31 (2): 1269- 1302.
35	KNUTH D . The art of computer programming[M]. 3rd ed MA: Addison Wesley, 1998.
36	DEKKER T J . A floating-point technique for extending the available precision[J]. Numerische Mathematik, 1971, 18 (3): 224- 242. doi: 10.1007/BF01397083

项目	浮点数类型
项目	半精度	单精度	双精度	四倍精度
字长/bits	16	32	64	128
符号位s/bits	1	1	1	1
指数位e/bits	5	8	11	15
精度位p/bits	11	24	53	113
上边界	-14	-126	-1 022	-16 382
下边界	15	127	1 023	16 383

算法	循环相位归零法	联合法
CudaMemcpyHostToDevice	34.8	34.5
DDCProcess	19.2	22.3
CudaMemcpyDeviceToHost	35.6	35.8

[1]	李冬, 焦义文, 高泽夫, 杨文革, 毛飞龙, 滕飞. 基于GPU的相位干涉仪FX鉴相算法[J]. 系统工程与电子技术, 2022, 44(11): 3320-3329.
[2]	曾艳阳，康凤举，许鹏，李小军. 水下战视景仿真系统建模及实时场景渲染方法[J]. Journal of Systems Engineering and Electronics, 2013, 35(2): 436-440.