IFFT_BOX


   more...


   more...

Computes the inverse Discrete Fourier Transform of several complex or real sequences.

Function Return Value

Complex array containing the inverse of the Discrete Fourier Transform of the sequences in X. If X is an assumed shape complex array of rank 2, 3 or 4, the result is a complex array of the same shape and rank consisting of the inverse DFT for each of the last rank’s indices.  (Output)

Required Argument

X — Box containing the sequences for which the inverse transform is to be computed. X is an assumed shape complex array of rank 2, 3 or 4. If X is real or double, it is converted to complex internally prior to the computation. (Input)

Optional Arguments, Packaged Options

WORK — A COMPLEX array of the same precision as the data. For rank-1 transforms the size of WORK is n + 15. To define this array for each problem, set WORK(1) = 0. Each additional rank adds the dimension of the transform plus 15. Using the optional argument WORK increases the efficiency of the transform.

The option and derived type names are given in the following tables:

Option Names for IFFT

Option Value

Options_for_fast_dft

1

 

Name of Unallocated Option Array to Use for Setting Options

Use

Derived Type

?_ifft_box_options(:)

Use when setting options for calls hereafter.

?_options

?_ifft_box_options_once(:)

Use when setting options for next call only.

?_options

For a description on how to use these options, see Matrix Optional Data Changes. See FAST_DFT located in Chapter 6, “Transforms” for the specific options for this routine.

FORTRAN 90 Interface

IFFT_BOX (X [])

Description

Computes the inverse of the Discrete Fourier Transform of a box of complex sequences. This function uses FAST_DFT, FAST_2DFT, and FAST_3DFT from Chapter 6.

Parallel Example

 

use rand_gen_int

use fft_box_int

use ifft_box_int

use linear_operators

use mpi_setup_int

 

implicit none

 

! This is FFT_BOX example.

 

integer i,j

integer, parameter :: n=40, nr=4

real(kind(1e0)) :: err(nr), one=1e0

real(kind(1e0)) :: a(n,1,nr), b(n,nr), c(n,1,nr), yy(n,n,nr)

complex(kind(1e0)), dimension(n,nr) :: f, fa, fb, cc, aa

 

real(kind(1e0)),parameter::zero_par=0.e0

real(kind(1e0))::dummy_par(0)

integer iseed_par

type(s_options)::iopti_par(2)

 

 

! setup for MPI

MP_NPROCS = MP_SETUP()

 

! Set Random Number generator seed

 

iseed_par = 53976279

iopti_par(1)=s_options(s_rand_gen_generator_seed,zero_par)

iopti_par(2)=s_options(iseed_par,zero_par)

 

call rand_gen(dummy_par,iopt=iopti_par)

 

! Generate two random periodic sequences 'a' and 'b'.

a=rand(a); b=rand(b)

 

! Compute the convolution 'c' of 'a' and 'b'.

do i=1,nr

aa(1:,i) = a(1:,1,i)

yy(1:,1,i)=b(1:,i)

do j=2,n

yy(2:,j,i)=yy(1:n-1,j-1,i)

yy(1,j,i)=yy(n,j-1,i)

end do

end do

 

c=yy .x. a

 

! Compute f=inverse(transform(a)*transform(b)).

fa = fft_box(aa)

fb = fft_box(b)

f=ifft_box(fa*fb)

 

! Check the Convolution Theorem:

! inverse(transform(a)*transform(b)) = convolution(a,b).

do i=1,nr

cc(1:,i) = c(1:,1,i)

end do

err = norm(cc-f)/norm(cc)

if (ALL(err <= sqrt(epsilon(one))) .AND. MP_RANK == 0) then

write (*,*) 'FFT_BOX is correct.'

end if

 

MP_NPROCS = MP_SETUP('Final')

end