Name

sfft_apply_grp, dfft_apply_grp, cfft_apply_grp, zfft_apply_grp − Application step for group fast fourier transform in one dimension

FORMAT

Real transform: status = {S,D}FFT_APPLY_GRP (input_format, output_format, direction, in, out, grp_size, lda, fft_struct, stride, grp_stride)

Complex transform of real data format: status = {C,Z}FFT_APPLY_GRP (input_format, output_format, direction, in, out, grp_size, lda, fft_struct, stride, grp_stride)

Complex transform of real data to real data: status = {C,Z}FFT_APPLY_GRP (input_format, output_format, direction, in_real, in_imag, out_real, out_imag, grp_size, lda, fft_struct, stride, grp_stride)

Arguments

input_format, output_format
character∗(∗)
Identifies the data type of the input and the format to be used to store the data, regardless of the data type.  For example, a complex sequence can be stored in real format.

The character ’R’ specifies the format as real; the character ’C’ specifies the format as complex. As convenient, use either upper- or lowercase characters, and either spell out or abbreviate the word. 

The following table shows the valid values:

For complex transforms, the type of data determines what other arguments are needed. When both the input and output data are real, the complex functions store the data as separate arrays for imaginary and real data so additional arguments are needed. 

directioncharacter∗(∗)
Specifies the operation as either the forward or inverse transform. Use ’F’ or ’f’ tp specify the forward transform. Use ’B’ or ’b’ to specify the the inverse transform.

in, outreal∗4 | real∗8 | complex∗8 | complex∗16
Both arguments are two-dimensional arrays. The IN array contains the data to be transformed. The OUT array contains the transformed data. The IN and OUT arrays can be the same array.

in_real, in_imag, out_real, out_imag
real∗4 | real∗8
Use these arguments when performing a complex transform on real data format and storing the result in a real data format.

grp_sizeinteger∗4
Specifies the number of one-dimensional data arrays; grp_size > 0. 

ldainteger∗4
Specifies the number of rows in two-dimensional data arrays; lda >= grp_size. Using lda = grp_size + {3 or 5} can sometimes achieve better performance by avoiding cache thrashing.

fft_structrecord /dxml_s_grp_fft_structure/ for single-precision real operations
record /dxml_d_grp_fft_structure/ for double-precision real operations
record /dxml_c_grp_fft_structure/ for single-precision complex operations
record /dxml_z_grp_fft_structure/ for double-precision complex operations
The argument refers to the structure created by the _INIT routine.

strideinteger∗4
Specifies the distance between columns of active data arrays; stride >= 1. stride permits columns of IN and OUT arrays to be skipped when they are not part of the group.

grp_strideinteger∗4
Specifies the distance between consecutive elements in a row in the IN and OUT arrays; grp_stride >= 1.  grp_stride permits rows of IN and OUT arrays to be skipped when they are not part of the group. 

Description

The _FFT_APPLY_GRP computes the fast forward or inverse Fourier transform on a group of one-dimensional data arrays.  The transform is performed on the first row of elements of one-dimensional data arrays within the group. Data array can be skipped by setting the stride parameter.  The transform then goes to the next row of elements. Similarly, rows of elements can be skipped by setting the grp_stride parameter.  _FFT_APPLY_GRP contrasts with _FFT_APPLY in that _FFT_APPLY perform a transform on each element of a data array before going to the next data array.  Although _FFT_APPLY_GRP gives the same result as _FFT_APPLY, _FFT_APPLY_GRP is more efficient at completing the transform. 

Return Values

0DXML_SUCCESS()

12DXML_INS_RES()

13DXML_BAD_STRIDE()

15DXML_BAD_DIRECTION_STRING()

16DXML_BAD_FORMAT_STRING()

18 (real transform only)
DXML_BAD_FORMAT_FOR_DIRECTION()

Example

INCLUDE {file_location}DXMLDEF.FOR
INTEGER∗4 GRP_SIZE,N,STATUS
REAL∗8 A(100,514),B(100,514)
RECORD /DXML_D_GRP_FFT_STRUCTURE/FFT_STRUCT
GRP_SIZE=100
N=512
LDA=100
STATUS = DFFT_INIT_GRP(N,FFT_STRUCT,.TRUE.,100)
STATUS = DFFT_APPLY_GRP(’R’,’C’,’F’,A,B,GRP_SIZE,LDA,FFT_STRUCT,1,1)
STATUS = DFFT_EXIT_GRP(FFT_STRUCT)

This FORTRAN code computes a set of 100 FFT of size 512. The results of the transforms are stored in B in complex format. Note that the second dimension is 514 to hold the extra complex values.