Changes between Version 16 and Version 17 of Examples/Fft2dHighpass

Show
Ignore:
Timestamp:
05/17/10 09:21:41 (5 years ago)
Author:
benl (IP: 129.94.242.47)
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • Examples/Fft2dHighpass

    v16 v17  
    4646|| Using FFTW using Estimate mode || 0.09 || [http://code.haskell.org/repa/repa-head/repa-examples/FFT/HighPass/legacy/c/FFTW.c FFTW.c] || 
    4747 
    48 The vector version uses the same recursive radix-2 decimation in time (DIT) algorithm as the Repa version, but is not rank generalised. It applies a recursive 1d FFT to each row and then transposes the matrix, twice each. Recursive FFT algorithms tend to be slower than in-place ones because the data is copied into new vectors at each recursion. A 512 point FFT is built from two 256 point FFTs, which are build from 4 128 point FFTs and so on. The result of each FFT is a new vector which needs to be allocated, filled, then unboxed again during the next recursion. 
     48The vector version uses the same recursive radix-2 decimation in time (DIT) algorithm as the Repa version, but is not rank generalised. It applies a recursive 1d FFT to each row and then transposes the matrix, twice each. Recursive FFT algorithms tend to be slower than in-place ones because the data is copied into new vectors at each recursion. A 512 point FFT is built from two 256 point FFTs, which are build from 4 128 point FFTs and so on. The result of each FFT is a new vector which needs to be allocated, filled, then unboxed again during the next recursion. The base case is a 2 point vector, so there are 256 of them to allocate then unbox at the lowest step. 
    4949 
    5050Jones's version also uses a 1d radix-2 DIT FFT kernel, but it first reorders the values then performs a in-place transform using three nested loops. Using an in-place algorithm gives better locality, and avoids the need to allocate and unbox all the intermediate vectors.