Accelerating batched 1D-FFT with a CUDA-capable computer
Department of Applied Physics, University of Technology Delft, Delft, Netherlands
IEEE International Conference on Imaging Systems and Techniques (IST), 2010
This work concerns the application of CUDA-based software (Compute Unified Device Architecture), developed by NVIDIA for programmable Graphics Processing units (GPUs). CUDA code is written in ‘C for CUDA’, indicating the standard C programming language with NVIDIA extensions.Our goal was to find out, whether batched (multiple) one-dimensional Fast Fourier Transformation (1DFFT), often encountered in various fields of signal processing, can be speeded up significantly by exploiting the parellel-processing power of a low-cost, standard, CUDA-enabled graphics card in a home-assembled PC.
June 20, 2011 by hgpu
