d@0: d@0: Next: Guru Real-data DFTs, d@0: Previous: Guru vector and transform sizes, d@0: Up: Guru Interface d@0:
fftw_plan fftw_plan_guru_dft( d@0: int rank, const fftw_iodim *dims, d@0: int howmany_rank, const fftw_iodim *howmany_dims, d@0: fftw_complex *in, fftw_complex *out, d@0: int sign, unsigned flags); d@0: d@0: fftw_plan fftw_plan_guru_split_dft( d@0: int rank, const fftw_iodim *dims, d@0: int howmany_rank, const fftw_iodim *howmany_dims, d@0: double *ri, double *ii, double *ro, double *io, d@0: unsigned flags); d@0:d@0:
d@0: These two functions plan a complex-data, multi-dimensional DFT
d@0: for the interleaved and split format, respectively.
d@0: Transform dimensions are given by (rank, dims) over a
d@0: multi-dimensional vector (loop) of dimensions (howmany_rank,
d@0: howmany_dims). dims and howmany_dims should point
d@0: to fftw_iodim arrays of length rank and
d@0: howmany_rank, respectively.
d@0:
d@0:
flags is a bitwise OR (`|') of zero or more planner flags,
d@0: as defined in Planner Flags.
d@0:
d@0:
In the fftw_plan_guru_dft function, the pointers in and
d@0: out point to the interleaved input and output arrays,
d@0: respectively. The sign can be either -1 (=
d@0: FFTW_FORWARD) or +1 (= FFTW_BACKWARD). If the
d@0: pointers are equal, the transform is in-place.
d@0:
d@0:
In the fftw_plan_guru_split_dft function,
d@0: ri and ii point to the real and imaginary input arrays,
d@0: and ro and io point to the real and imaginary output
d@0: arrays. The input and output pointers may be the same, indicating an
d@0: in-place transform. For example, for fftw_complex pointers
d@0: in and out, the corresponding parameters are:
d@0:
d@0:
ri = (double *) in; d@0: ii = (double *) in + 1; d@0: ro = (double *) out; d@0: io = (double *) out + 1; d@0:d@0:
Because fftw_plan_guru_split_dft accepts split arrays, strides
d@0: are expressed in units of double. For a contiguous
d@0: fftw_complex array, the overall stride of the transform should
d@0: be 2, the distance between consecutive real parts or between
d@0: consecutive imaginary parts; see Guru vector and transform sizes. Note that the dimension strides are applied equally to the
d@0: real and imaginary parts; real and imaginary arrays with different
d@0: strides are not supported.
d@0:
d@0:
There is no sign parameter in fftw_plan_guru_split_dft.
d@0: This function always plans for an FFTW_FORWARD transform. To
d@0: plan for an FFTW_BACKWARD transform, you can exploit the
d@0: identity that the backwards DFT is equal to the forwards DFT with the
d@0: real and imaginary parts swapped. For example, in the case of the
d@0: fftw_complex arrays above, the FFTW_BACKWARD transform
d@0: is computed by the parameters:
d@0:
d@0:
ri = (double *) in + 1; d@0: ii = (double *) in; d@0: ro = (double *) out + 1; d@0: io = (double *) out; d@0:d@0: d@0: d@0: