一种支持多数据块混合处理的FFT优化方法

doi:10.19665/j.issn1001-2400.2022.06.006

摘要/Abstract

摘要：

针对快速傅里叶变换处理器中运算通路深流水线气泡会导致性能损失以及不同点数的快速傅里叶变换存在吞吐率不均衡问题,提出了一种可以同时支持多个快速傅里叶变换数据混合处理的优化方法。设计了一种深度流水的可配置蝶形处理电路以及可支持多数据块混合处理的块浮点处理架构,使得在同一硬件框架下可支持1个基9/2个基8/3个基5/4个基4/5个基3的高精度蝶形运算,运算速率和资源效率较高。基于上述方法,实现了一种支持4G/5G标准的多模高性能快速傅里叶变换处理器,可以支持64~4 096 点的FFT/iFFT和12~3 240点的DFT/iDFT处理(60种点数模式)。该快速傅里叶变换处理器基于55 nm CMOS工艺实现,面积1.59 mm²,最高工作频率500 MHz,单数据模式下最大吞吐率 1.5 GS/s,混合数据模式下最大吞吐率2.2 GS/s。与近年研究相比,该设计在增加较少资源的情况下,实现了更多点数支持、更高的吞吐率(2X~6X)和各种点数下更均衡的性能。

关键词: 快速傅里叶变换处理器, 高性能, 多模式, 5G

Abstract:

A new computational method is proposed to support multiple FFT processing simultaneously.The method can effectively solve the performance loss of memory-based architecture FFT processors due to the computational path pipeline bubbles and the unbalanced throughput of different FFT points.A deeply pipelined WFTA algorithm-based configurable butterfly unit and a new multiple block floating-point processing unit are designed,which can support high precision computing and include one radix-9/two radix-8/three radix-5/four radix-4/five radix-3 in parallel.Based on the proposed method,a multi-mode high performance FFT processor for 4G/5G is implemented,which can support 60 modes including 64~4 096-point FFT/iFFT and 12~3 240-point DFT/iDFT computing.The processor is implemented based on 55 nm CMOS technology,with a 1.46 mm² layout,supports 1.5GS/s in a single mode and 2.2 GS/s in a mixed mode at 500MHz.It is shown that the proposed processor can support more points and has a better performance than previous designs.

Key words: FFT processor, high-performance, multi-mode, 5G

中图分类号:

TN432

洪钦智,王志君,郭一凡,梁利平. 一种支持多数据块混合处理的FFT优化方法[J]. 西安电子科技大学学报, 2022, 49(6): 42-50.

HONG Qinzhi,WANG Zhijun,GUO Yifan,LIANG Liping. Multi-data mixed FFT processing optimization method[J]. Journal of Xidian University, 2022, 49(6): 42-50.

图/表 13

图1

图2

表1

表2

图3

表3

图4

图5

图6

图7

表4

表5

表6

参考文献 14

[1]	SESIA S, TOUFIK I, BAKER M. LTE, The UMTS Long Term Evolution:From Theory to Practice[M]. Hoboken: Wiley Publishing, 2009:1-591.
[2]	LAN/MAN Standards Committee. Specific Requirements Part 11:Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications Amendment 4:Enhancements for Very High Throughput for Operation in Bands below 6 GHz IEEE Computer Society:IEEE Std 802.11ac-2013[S]. Piscataway:IEEE, 2013.
[3]	PANWAR N, SHARMA S, SINGH A K. A Survey on 5G:The Next Generation of Mobile Communication[J]. Physical Communication, 2016, 18(2):64-84. doi: 10.1016/j.phycom.2015.10.006
[4]	KONGUVEL E, KANNAN M. A Survey on FFT/IFFT Processors for Next Generation Telecommunication Systems[J]. Journal of Circuits Systems & Computers, 2017, 27(3):1830001
[5]	SHAHABUDDIN S, MAMMELA A, JUNTTI M, et al. ASIP for 5G and Beyond:Opportunities and Vision[J]. IEEE Transactions on Circuits and Systems II:Express Briefs, 2021, 68(3):851-857. doi: 10.1109/TCSII.2021.3050785
[6]	白宝明, 马啸, 陈文, 等. 面向B5G/6G的信息传输与接入技术专题序言[J]. 西安电子科技大学学报, 2020, 47(6):1-4.
	BAI Baoming, MA Xiao, CHEN Wen, et al. Preface to The Topic of Information Transmission and Access Technology for B5G/6G[J]. Journal of Xidian university, 2020, 47(6):1-4.
[7]	MIHRET E T, HAILEG . 4G,5G,6G,7G and Future Mobile Technologies[J]. American Journal of Computer Science and Technology, 2021, 9(2):75.
[8]	杨瑞科, 高霞, 武福平, 等. 5G毫米波雨衰减链路传播模型及通信性能[J]. 西安电子科技大学学报, 2022, 49(4):24-30.
	YANG Ruike, GAO Xia, WU Fuping, et al. Research on The Propagation Model and Communication Performance of the 5G Millimeter Wave at The Rain Attenuation Link[J]. Journal of Xidian University, 2022, 49(4):24-30.
[9]	PATEL D K, SINGH S. VLSI Architecture for Combined R2B,R4B and R8B FFT Using SDF and Modified CSLA[C]//2021 International Conference on Computer Communication and Informatics (ICCCI). Piscataway:IEEE, 2021:1-5.
[10]	SHIH X Y, CHOU H R, LIU Y Q. VLSI Design and Implementation of Reconfigurable 46-Mode Combined-Radix-Based FFT Hardware Architecture for 3GPP-LTE Applications[J]. IEEE Transactions on Circuits and Systems I:Regular Papers, 2018, 65(1):118-129. doi: 10.1109/TCSI.2017.2725338
[11]	CHEN J, HU J, LEE S, et al. Hardware Efficient Mixed Radix-25/16/9 FFT for LTE Systems[J]. IEEE Transactions on Very Large Scale Integration Systems, 2015, 23(2):221-229. doi: 10.1109/TVLSI.2014.2304834
[12]	XIA K F, WU B, XIONG T, et al. A Memory-Based FFT Processor Design with Generalized Efficient Conflict-Free Address Schemes[J]. IEEE Transactions on Very Large Scale Integration Systems, 2017, 25(6):1919-1929. doi: 10.1109/TVLSI.2017.2666820
[13]	LIU S, LIU D. A High-Flexible Low-Latency Memory-Based FFT Processor for 4G,WLAN,and Future 5G[J]. IEEE Transactions on Very Large Scale Integration (VLSI) Systems, 2019, 27(3):511-523. doi: 10.1109/TVLSI.2018.2879675
[14]	SILVERMAN H F. An Introduction to Programming the Winograd Fourier Trans form Algorithm (WFTA)[J]. IEEE Trans on, 1977, 25(2):152-165.

基数	并行度	乘法器数	加法器数
Radix3	5	2×5=10	6×5=30
Radix4	4	0×4=0	8×4=32
Radix5	3	4×3=12	17×3=51
Radix8	2	2×2=4	26×2=52
Radix9	1	10×1=10	42×1=42

文献	乘法	加法	Radix	等效计算周期
文献	乘法	加法	Radix	R3	R4	R5	R8	R9
文中算法	12	52	3/4/5/8/9	0.2	0.25	0.33	0.5	1
文献[13]	8	66	2/4/8/16/3/5	0.25	0.25	0.5	0.5	2
文献[11]	10	34	2/3/4/5/8/9/16/25	0.5	1	1	2	3

FFT运算点数类型	支持并行度
12,24,36,48,60,64	64
72,96,108,120,128	32
144,180,192,216,240,256	16
288,300,324,360,384,432,480,512	8
540,576,600,648,720,768,864,900,960,972,1 024	4
1 080,1 152,1 200,1 296,1 440,1 500,1 536,1 620,1 728,1 800,1 920,1 944,2 048	2
2 160,2 304,2 400,2 592,2 700,2 880,2 916,3 000,3 072,3 240,4 096	1

点数	文中	文中REPEAT模式			文献 [13]	文献 [11]
点数	文中	并行度	AC	eAC	文献 [13]	文献 [11]
12	27	64	200	4	23	22
144	78	16	738	47	72	72
216	100	16	1 082	68		173
384	178	8	1 116	140	134	267
480	210	8	1 372	172	226	427
540	248	4	859	215		524
648	279	4	1 012	253	400	605
864	388	4	1 420	355	484	875
972	457	4	1 695	424	671	1 064
1 080	448	2	851	426	688	1 037
1 200	585	2	1 116	558	613	751
1 296	607	2	1 170	585	733	1 199
2 048	696	2	1 348	674	542
4 096	1332	1	1 332	1 332	1 054

点数	文中		文献 [13]	文献 [11]
点数	单数据模式	REPEAT模式	文献 [13]	文献 [11]
12	222	1 500	130	67
144	923	1 532	500	246
216	1 080	1 588		153
384	1 079	1 371	716	177
480	1 143	1 395	531	138
540	1 089	1 256		127
648	1 161	1 281	405	132
864	1 113	1 217	446	121
972	1 063	1 146	362	112
1 080	1 205	1 268	392	128
1 200	1 026	1 075	489	196
1 296	1 068	1 108	442	133
2 048	1 471	1 519	945
4 096	1 538	1 538	972