• DocumentCode
    1777621
  • Title

    Vector implementation of the fast Fourier transform on DSP and NVIDIA CUDA platforms

  • Author

    Pikacz, Bartosz ; Gambrych, Jacek

  • Author_Institution
    Warsaw Univ. of Technol., Warsaw, Poland
  • fYear
    2014
  • fDate
    June 30 2014-July 3 2014
  • Firstpage
    1
  • Lastpage
    4
  • Abstract
    This paper presents two implementations of the fast Fourier transform decomposed into vector operations. This approach is appropriate for cases where data to be transformed is stored in an unorthodox order, and as such is well suited for radar and sonar data processing. The described procedures performing a vector FFT were implemented for TigerSHARC DSP and NVIDIA CUDA platforms. Their performance was measured and compared with highly optimized library procedures.
  • Keywords
    digital signal processing chips; fast Fourier transforms; parallel architectures; NVIDIA CUDA platforms; TigerSHARC DSP; fast Fourier transform; optimized library procedures; radar; sonar data processing; unorthodox order; vector FFT; Digital signal processing; Libraries; Performance evaluation; Transforms; Vectors;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Ph.D. Research in Microelectronics and Electronics (PRIME), 2014 10th Conference on
  • Conference_Location
    Grenoble
  • Type

    conf

  • DOI
    10.1109/PRIME.2014.6872729
  • Filename
    6872729