#include <optFFT_API.h>
Public Member Functions | |
| virtual void | optimizedFFT (FFT_precision **&Real, FFT_precision **&Imag, const int size_x, const int size_y)=0 |
| virtual void | optimizedIFFT (FFT_precision **&Real, FFT_precision **&Imag, const int size_x, const int size_y)=0 |
| virtual | ~optFFT_API () |
| virtual destructor for inheritance | |
Detailed Description
API of the optimized FFT
Perform all operations in place
Member Function Documentation
◆ optimizedFFT()
|
pure virtual |
Perform an optimized FFT using the symmetry of the expansions
Assume that the matrices are the real and imag part of the expansion for Expansion(l,m), l is the first coordinate of the matrices The symmetry Expansion(l,-m) = (-1)^m conj(Expansion(l,m)) is valid
The resulting FFT is done on size_x,2*size_y matrices that would represent on the line i Ex(i,0),.., Ex(i,size_x), | 0, Ex(i,-size_x),.., Ex(i,-1) (the part after | is reconstruct for the FFT and doesn't appear outside of it or in the result, Ex(i,j)=0 if j<i or j>i)
The output is the first half of the reconstructed matrix since it verifies M(i,j) = conj(M(i,j+size_y)) The output is safe to use for entrywise default operation with other matrices converted with this function
- Parameters
-
FFT_precision** & Real, real part FFT_precision** & Imag, imag part const int size_x, size of the first coord of the matrices const int size_y, size of the second coord of the matrices
Implemented in FakedOptFFT.
◆ optimizedIFFT()
|
pure virtual |
Inverse FFT Result is assumed to be rescaled
Take a matrix couples converted with optimizedFFT as input and assume the symmetry M(i,j) = conj(M(i,j+size_y)) on the reconstructed full matrix
Output the first half of the reconstructed matrix IFFT
Implemented in FakedOptFFT.
The documentation for this class was generated from the following file:
- src/bhfmm/fft/tools/optimizedFFT/optFFT_API.h
