Journal of South China University of Technology(Natural Science) >

2025 , Vol. 53 >Issue 3: 20 - 30

DOI: https://doi.org/10.12141/j.issn.1000-565X.240035

Computer Science & Technology

Design and Optimization of High-Performance Multi-Dimensional FFT Based on Matrix Core

  • LU Lu ,
  • ZHU Songxiang ,
  • TIAN Qingyan ,
  • LIN Haishan ,
  • GUO Yijie
Expand
  • 1.School of Computer Science and Engineering，South China University of Technology，Guangzhou 510006，Guangdong，China
    2.Pengcheng Laboratory，Shenzhen 518000，Guangdong，China
    3.Tunnel Engineering Safety and Emergency Support Technology and Equipment Laboratory of Guangdong Province，Guangzhou 510440，Guangdong，China
陆璐（1971—），男，博士，教授，主要从事软件缺陷预测、高性能计算、深度学习训练推理加速研究。E-mail： lul@scut.edu.cn

Received date: 2024-01-12

  Online published: 2024-04-23

Supported by

the Key-Area R & D Program of Guangdong Province(2022B0101070001)

Fold

Abstract

Fast Fourier transform (FFT) algorithm finds widespread application in scientific computing and related fields. To fully leverage the computational power of the GPU and further enhance the performance of FFT calculations, this paper proposed a high-performance multi-dimensional FFT computation scheme based on the Matrix Core for the matrix form of Stockham FFT. In terms of computational optimization, this scheme utilizes Matrix Core to accelerate matrix multiplications in FFT computation while leveraging compiler intrinsic instructions to perform small-grained matrix multiply-accumulate operations, enabling Matrix Core to support FFT computations of more sizes. To minimize memory access, the proposed scheme directly performs matrix element-wise multiplication operations in the registers according to the distribution pattern of Matrix Core’s data across thread registers. It also mitigates bank conflicts by reordering data in shared memory, adopts a double-buffering strategy to alleviate access bottlenecks, and proposes an efficient matrix transposition strategy to accelerate multidimensional FFT computations. In this paper, the proposed scheme was compared to the well-known high-performance FFT computation libraries rocFFT and VkFFT on the AMD MI250 GPU platform. The results demonstrate that the proposed scheme outperforms rocFFT and VkFFT in terms of average computational performance for 1-dimensional, 2-dimensional, and 3-dimensional FFTs on the AMD MI250 GPU platform. For 3D FFT calculation, this method has an average performance that is 1.5 times faster than rocFFT and 2.0 times faster than VkFFT, demonstrating significant performance improvements.

Key words： graphics processing unit; Matrix Core; fast Fourier transform; matrix multiplication

Cite this article

LU Lu , ZHU Songxiang , TIAN Qingyan , LIN Haishan , GUO Yijie . Design and Optimization of High-Performance Multi-Dimensional FFT Based on Matrix Core[J]. Journal of South China University of Technology(Natural Science), 2025 , 53(3) : 20 -30 . DOI: 10.12141/j.issn.1000-565X.240035

References

1 JUNG J， KOBAYASHI C， IMAMURA T，et al ．Parallel implementation of 3D FFT with volumetric decomposition schemes for efficient molecular dynamics simulations［J］．Computer Physics Communications，2016，200：57-65.
2 LOGESHWARAN N， VIJAYAPRADEEP S， KIM A R，et al ．Study of engineering electronic structure modulated non-noble metal oxides for scaled-up alkaline blend seawater splitting［J］．Journal of Energy Chemistry，2023，86：167-179.
3 MOHAPATRA B N， MOHAPATRA R K ．FFT and sparse FFT techniques and applications［C］∥ Procee-dings of 2017 the Fourteenth International Conference on Wireless and Optical Communications Networks．Mumbai：IEEE，2017：1-5.
4 DESPRéS P， JIA X ．A review of GPU-based medical image reconstruction［J］．Physica Medica，2017，42：76-92.
5 CHI L， JIANG B， MU Y ．Fast Fourier convolution［C］∥ Advances in Neural Information Processing Systems 33：34th Conference on Advances in Neural Information Processing Systems．San Diego：Neural Information Proce-ssing Systems Foundation，Inc.，2020：4479-4488.
6 LUSZCZEK P， DONGARRA J J， KOESTER D，et al ．Introduction to the HPC challenge benchmark suite［R］．Denton：UNT Libraries Government Documents Department，2005：1-12.
7 LELIAERT J， MULKERS J ．Tomorrow’s micromagnetic simulations［J］．Journal of Applied Physics，2019，125（18）：180901/1-9.
8 MILOJICIC D ．Accelerators for artificial intelligence and high-performance computing［J］．Computer，2020， 53（2）：14-22.
9 TOLMACHEV D ．VkFFT：a performant，cross-platform and open-source GPU FFT library［J］．IEEE Access，2023，11：12039-12058.
10 ZHANG F，HU　C，YIN　Q，et al ．A GPU based memory optimized parallel method for FFT implementation［EB/OL］．（2017-07-23） ［2024-01-12］． ．
11 LEE J，KIM　D ．Large-scale 3D fast Fourier transform computation on a GPU［J］．ETRI Journal，2023，45（6）：1035-1045.
12 ZHAO Y， LIU F， MA W，et al ．MFFT：a GPU accelerated highly efficient mixed-precision large-scale FFT framework［J］．ACM Transactions on Architecture and Code Optimization，2023，20（3）：1-23.
13 NVIDIA Corporation ．NVIDIA Tesla V100 GPU architecture［EB/OL］． （2017-08-01）［2024-01-12］．.
14 SORNA A， CHENG X， D’AZEVEDO E，et al ．Optimizing the fast Fourier transform using mixed precision on tensor core hardware［C］∥ Proceedings of IEEE the 25th International Conference on High Performance Computing Workshops．Bengaluru：IEEE，2018：3-7.
15 DURRANI S， CHUGHTAI M S， HIDAYETOGLU M，et al ．Accelerating Fourier and number theoretic transforms using tensor cores and warp shuffles［C］∥ Proceedings of the 30th International Conference on Para-llel Architectures and Compilation Techniques．Atlanta：IEEE，2021：345-355.
16 PISHA L， LIGOWSKI L ．Accelerating non-power-of-2 size FOURIER transforms with GPU tensor cores［C］∥ Proceedings of 2021 IEEE International Parallel and Distributed Processing Symposium．Portland：IEEE，2021：507-516.
17 LI B， CHENG S， LIN J ．tcFFT：a fast half-precision FFT library for NVIDIA tensor cores［C］∥ Proceedings of 2021 IEEE International Conference on Cluster Compu-ting．Portland：IEEE，2021：1-11.
18 TUKEY C， COOLEY J ．An algorithm for the machine calculation of complex Fourier series［J］．Mathema-tics of Computation，1965，19（90）：297-301.
19 Advanced Micro Devices Inc． CDNA Introducing AMD 2 architecture［EB/OL］．（2021-09-21）［2024-01-12］．.
20 FRIGO M， JOHNSON S ．FFTW：an adaptive software architecture for the FFT［C］∥ Proceedings of the 1998 IEEE International Conference on Acoustics，Speech and Signal Processing．Seattle：IEEE，1998：1381-1384.
Options
Outlines

/

〈 〉