Next Page >

Hide TOC

vDSP Two-Dimensional Fast Fourier Transforms Reference

Framework	Accelerate/vecLib
Declared in	vDSP.h

Overview

Describes the C API for the functionality for two-dimensional Fast Fourier Transforms in vecLib.

Functions by Task

Computing In-Place Complex FFTs

Computing Out-of-Place Complex FFTs

Computing In-Place Real FFTs

Computing Out-of-Place Real FFTs

Functions

vDSP_fft2d_zip

Computes an in-place single-precision complex discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse).

   void vDSP_fft2d_zip (FFTSetup setup,
   DSPSplitComplex * ioData,
   vDSP_Stride strideInRow,
   vDSP_Stride strideInCol,
   vDSP_Length log2nInCol,
   vDSP_Length log2nInRow,
   FFTDirection direction);

Parameters

setup

Points to a structure initialized by a prior call to FFT weights array function vDSP_create_fftsetup or vDSP_create_fftsetupD. The value supplied as parameter log2n of the setup function must equal or exceed the values supplied as parameters log2nInCol and log2nInRow of the transform function.

ioData

A complex vector input.

strideInRow

Specifies a stride across each row of the matrix signal. Specifying 1 for strideInRow processes every element across each row, specifying 2 processes every other element across each row, and so forth.

strideInCol

Specifies a column stride for the matrix, and should generally be allowed to default unless the matrix is a submatrix. Parameter strideInCol can be defaulted by specifying 0. The default column stride equals the row stride multiplied by the column count. Thus, if strideInRow is 1 and strideInCol is 0, every element of the input /output matrix is processed. If strideInRow is 2 and strideInCol is 0, every other element of each row is processed.

If not 0, parameter strideInCol represents the distance between each row of the matrix. If parameter strideInCol is 1024, for instance, complex element 512 of the matrix equates to element (1,0), element 1024 equates to element (2,0), and so forth.

log2nInCol

The base 2 exponent of the number of columns to process for each row. log2nInCol must be between 2 and 10, inclusive.

log2nInRow

The base 2 exponent of the number of rows to process. For example, to process 64 rows of 128 columns, specify 7 for parameter log2nInCol and 6 for parameter log2nInRow. log2nInRow must be between 2 and 10, inclusive.

direction

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of direction.

Discussion

This performs the operation

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_zipD

Computes an in-place double-precision complex discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse).

   void vDSP_fft2d_zipD (FFTSetupD setup,
   DSPDoubleSplitComplex * ioData,
   vDSP_Stride strideInRow,
   vDSP_Stride strideInCol,
   vDSP_Length log2nInCol,
   vDSP_Length log2nInRow,
   FFTDirection direction);

Parameters

setup

ioData

A complex vector input.

strideInRow

strideInCol

log2nInCol

The base 2 exponent of the number of columns to process for each row. log2nInCol must be between 2 and 10, inclusive.

log2nInRow

direction

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of direction.

Discussion

This performs the operation

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_zipt

Computes an in-place single-precision complex discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse). A buffer is used for intermediate results.

   void vDSP_fft2d_zipt (FFTSetup setup,
   DSPSplitComplex * ioData,
   vDSP_Stride strideInRow,
   vDSP_Stride strideInCol,
   DSPSplitComplex * bufferTemp,
   vDSP_Length log2nInCol,
   vDSP_Length log2nInRow,
   FFTDirection direction);

Parameters

setup

ioData

A complex vector input.

strideInRow

strideInCol

bufferTemp

A temporary matrix used for storing interim results. The size of temporary memory for each part (real and imaginary) is the lower value of 16 KB or 4*n, where log2n = log2nInCol + log2nInRow.

log2nInCol

The base 2 exponent of the number of columns to process for each row. log2nInCol must be between 2 and 10, inclusive.

log2nInRow

direction

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of direction.

Discussion

This performs the operation

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_ziptD

Computes an in-place double-precision complex discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse). A buffer is used for intermediate results.

   void vDSP_fft2d_ziptD (FFTSetupD setup,
   DSPDoubleSplitComplex * ioData,
   vDSP_Stride strideInRow,
   vDSP_Stride strideInCol,
   DSPDoubleSplitComplex * bufferTemp,
   vDSP_Length log2nInCol,
   vDSP_Length log2nInRow,
   FFTDirection direction);

Parameters

setup

ioData

A complex vector input.

strideInRow

strideInCol

bufferTemp

A temporary matrix used for storing interim results. The size of temporary memory for each part (real and imaginary) is the lower value of 16 KB or 4*n, where log2n = log2nInCol + log2nInRow.

log2nInCol

The base 2 exponent of the number of columns to process for each row. log2nInCol must be between 2 and 10, inclusive.

log2nInRow

direction

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of direction.

Discussion

This performs the operation

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_zop

Computes an out-of-place single-precision complex discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse).

   void vDSP_fft2d_zop (FFTSetup setup,
   DSPSplitComplex * signal,
   vDSP_Stride signalStrideInRow,
   vDSP_Stride signalStrideInCol,
   DSPSplitComplex * result,
   vDSP_Stride strideResultInRow,
   vDSP_Stride strideResultInCol,
   vDSP_Length log2nInCol,
   vDSP_Length log2nInRow,
   FFTDirection flag);

Parameters

setup

signal

A complex vector signal input.

signalStrideInRow

Specifies a stride across each row of matrix a. Specifying 1 for signalStrideInRow processes every element across each row, specifying 2 processes every other element across each row, and so forth.

signalStrideInCol

If not 0, this parameter represents the distance between each row of the input /output matrix.

result

The complex vector signal output.

strideResultInRow

Specifies a row stride for output matrix result in the same way that signalStrideInRow specifies a stride for input the input /output matrix.

strideResultInCol

Specifies a column stride for output matrix result in the same way that signalStrideInCol specifies a stride for input the input /output matrix.

log2nInCol

The base 2 exponent of the number of columns to process for each row. log2nInCol must be between 2 and 10, inclusive.

log2nInRow

The base 2 exponent of the number of rows to process. For example, to process 64 rows of 128 columns, specify 7 for log2nInCol and 6 for log2nInRow. log2nInRow must be between 2 and 10, inclusive.

flag

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of flag.

Discussion

This performs the operation

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_zopD

Computes an out-of-place double-precision complex discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse).

   void vDSP_fft2d_zopD (FFTSetupD setup,
   DSPDoubleSplitComplex * signal,
   vDSP_Stride signalStrideInRow,
   vDSP_Stride signalStrideInCol,
   DSPDoubleSplitComplex * result,
   vDSP_Stride strideResultInRow,
   vDSP_Stride strideResultInCol,
   vDSP_Length log2nInCol,
   vDSP_Length log2nInRow,
   FFTDirection flag);

Parameters

setup

signal

A complex vector signal input.

signalStrideInRow

signalStrideInCol

If not 0, this parameter represents the distance between each row of the input /output matrix.

result

The complex vector signal output.

strideResultInRow

Specifies a row stride for output matrix result in the same way that signalStrideInRow specifies a stride for input the input /output matrix.

strideResultInCol

Specifies a column stride for output matrix result in the same way that signalStrideInCol specifies a stride for input the input /output matrix.

log2nInCol

The base 2 exponent of the number of columns to process for each row. log2nInCol must be between 2 and 10, inclusive.

log2nInRow

flag

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of flag.

Discussion

This performs the operation

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_zopt

Computes an out-of-place single-precision complex discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse). A buffer is used for intermediate results.

   void vDSP_fft2d_zopt (FFTSetup setup,
   DSPSplitComplex * signal,
   vDSP_Stride signalStrideInRow,
   vDSP_Stride signalStrideInCol,
   DSPSplitComplex * result,
   vDSP_Stride strideResultInRow,
   vDSP_Stride strideResultInCol,
   DSPSplitComplex * bufferTemp,
   vDSP_Length log2nInCol,
   vDSP_Length log2nInRow,
   FFTDirection flag);

Parameters

setup

signal

A complex vector signal input.

signalStrideInRow

signalStrideInCol

If not 0, this parameter represents the distance between each row of the input /output matrix. If parameter signalStrideInCol is 1024, for instance, element 512 equates to element (1,0) of matrix a, element 1024 equates to element (2,0), and so forth.

result

The complex vector signal output.

strideResultInRow

Specifies a row stride for output matrix result in the same way that signalStrideInRow specifies a stride for input the input /output matrix.

strideResultInCol

Specifies a column stride for output matrix result in the same way that signalStrideInCol specifies a stride for input the input /output matrix.

bufferTemp

A temporary matrix used for storing interim results. The size of temporary memory for each part (real and imaginary) is the lower value of 16 KB or 4*n, where log2n = log2nInCol + log2nInRow.

log2nInCol

The base 2 exponent of the number of columns to process for each row. log2nInCol must be between 2 and 10, inclusive.

log2nInRow

flag

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of flag.

Discussion

This performs the operation

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_zoptD

Computes an out-of-place double-precision complex discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse). A buffer is used for intermediate results.

   void vDSP_fft2d_zoptD (FFTSetupD setup,
   DSPDoubleSplitComplex * signal,
   vDSP_Stride signalStrideInRow,
   vDSP_Stride signalStrideInCol,
   DSPDoubleSplitComplex * result,
   vDSP_Stride strideResultInRow,
   vDSP_Stride strideResultInCol,
   DSPDoubleSplitComplex * bufferTemp,
   vDSP_Length log2nInCol,
   vDSP_Length log2nInRow,
   FFTDirection flag);

Parameters

setup

signal

A complex vector signal input.

signalStrideInRow

signalStrideInCol

result

The complex vector signal output.

strideResultInRow

Specifies a row stride for output matrix result in the same way that signalStrideInRow specifies a stride for input the input /output matrix.

strideResultInCol

Specifies a column stride for output matrix result in the same way that signalStrideInCol specifies a stride for input the input /output matrix.

bufferTemp

A temporary matrix used for storing interim results. The size of temporary memory for each part (real and imaginary) is the lower value of 16 KB or 4*n, where log2n = log2nInCol + log2nInRow.

log2nInCol

The base 2 exponent of the number of columns to process for each row. log2nInCol must be between 2 and 10, inclusive.

log2nInRow

flag

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of flag.

Discussion

This performs the operation

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_zrip

Computes an in-place single-precision real discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse).

   void vDSP_fft2d_zrip (FFTSetup setup,
   DSPSplitComplex * ioData,
   vDSP_Stride strideInRow,
   vDSP_Stride strideInCol,
   vDSP_Length log2nInCol,
   vDSP_Length log2nInRow,
   FFTDirection direction);

Parameters

setup

Points to a structure initialized by a prior call to FFT weights array function vDSP_create_fftsetup or vDSP_create_fftsetupD. The value supplied as parameter log2n of the setup function must equal or exceed the values supplied as parameters log2nInCol and log2nInRow of the transform function.

ioData

A complex vector input.

strideInRow

Specifies a stride across each row of the input matrix signal. Specifying 1 for strideInRow processes every element across each row, specifying 2 processes every other element across each row, and so forth.

strideInCol

If not 0, strideInCol represents the distance between each row of the matrix. If strideInCol is 1024, for instance, complex element 512 of the matrix equates to element (1,0), element 1024 equates to element (2,0), and so forth.

log2nInCol

The base 2 exponent of the number of columns to process for each row. log2nInCol must be between 2 and 10, inclusive.

log2nInRow

direction

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of direction.

Discussion

Forward transforms read real input and write packed complex output. Inverse transforms read packed complex input and write real output. As a result of packing the frequency-domain data, spatial-domain data and its equivalent frequency-domain data have the same storage requirements.

Real data is stored in split complex form, with odd reals stored on the imaginary side of the split complex form and even reals in stored on the real side.

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_zripD

Computes an in-place double-precision real discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse).

   void vDSP_fft2d_zripD (FFTSetupD setup,
   DSPDoubleSplitComplex * signal,
   vDSP_Stride strideInRow,
   vDSP_Stride strideInCol,
   vDSP_Length log2nInCol,
   vDSP_Length log2nInRow,
   FFTDirection flag);

Parameters

setup

Points to a structure initialized by a prior call to FFT weights array function vDSP_create_fftsetup or vDSP_create_fftsetupD. The value supplied as parameter log2n of the setup function must equal or exceed the values supplied as parameters log2nInCol and log2nInRow of the transform function.

signal

A complex vector signal input.

strideInRow

strideInCol

log2nInCol

The base 2 exponent of the number of columns to process for each row. log2nInCol must be between 2 and 10, inclusive.

log2nInRow

flag

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of flag.

Discussion

Real data is stored in split complex form, with odd reals stored on the imaginary side of the split complex form and even reals in stored on the real side.

No Altivec/SSE support for double precision. The function always invokes scalar code.

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_zript

Computes an in-place single-precision real discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse). A buffer is used for intermediate results.

   void vDSP_fft2d_zript (FFTSetup setup,
   DSPSplitComplex * ioData,
   vDSP_Stride strideInRow,
   vDSP_Stride strideInCol,
   DSPSplitComplex * bufferTemp,
   vDSP_Length log2nInCol,
   vDSP_Length log2nInRow,
   FFTDirection direction);

Parameters

setup

Points to a structure initialized by a prior call to FFT weights array function vDSP_create_fftsetup or vDSP_create_fftsetupD. The value supplied as parameter log2n of the setup function must equal or exceed the values supplied as parameters log2nInCol and log2nInRow of the transform function.

ioData

A complex vector input.

strideInRow

strideInCol

bufferTemp

A temporary matrix used for storing interim results. The size of temporary memory required is discussed below.

log2nInCol

The base 2 exponent of the number of columns to process for each row. log2nInCol must be between 3 and 10, inclusive.

log2nInRow

direction

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of direction.

Discussion

Real data is stored in split complex form, with odd reals stored on the imaginary side of the split complex form and even reals in stored on the real side.

The space needed in bufferTemp is at most max(9*nr, nc/2) elements in each of realp and imagp. Here is an example of how to allocate the space:

 int nr, nc, tempSize;

 nr = 1<<log2InRow;

 nc = 1<<log2InCol;

 tempSize = max(9*nr, nc/2);

 bufferTemp.realp = ( float* ) malloc (tempSize * sizeOf ( float ) );

 bufferTemp.imagp = ( float* ) malloc (tempSize * sizeOf ( float ) );

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_zriptD

Computes an in-place double-precision real discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse). A buffer is used for intermediate results.

   void vDSP_fft2d_zriptD (FFTSetupD setup,
   DSPDoubleSplitComplex * signal,
   vDSP_Stride strideInRow,
   vDSP_Stride strideInCol,
   DSPDoubleSplitComplex * bufferTemp,
   vDSP_Length log2nInCol,
   vDSP_Length log2nInRow,
   FFTDirection flag);

Parameters

setup

Points to a structure initialized by a prior call to FFT weights array function vDSP_create_fftsetup or vDSP_create_fftsetupD. The value supplied as parameter log2n of the setup function must equal or exceed the values supplied as parameters log2nInCol and log2nInRow of the transform function.

signal

A complex vector signal input.

strideInRow

strideInCol

bufferTemp

A temporary matrix used for storing interim results. The size of temporary memory required is discussed below.

log2nInCol

The base 2 exponent of the number of columns to process for each row. log2nInCol must be between 3 and 10, inclusive.

log2nInRow

flag

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of flag.

Discussion

Real data is stored in split complex form, with odd reals stored on the imaginary side of the split complex form and even reals in stored on the real side.

The space needed in bufferTemp is at most max(9*nr, nc/2) elements in each of realp and imagp. Here is an example of how to allocate the space:

 int nr, nc, tempSize;

 nr = 1<<log2InRow;

 nc = 1<<log2InCol;

 tempSize = max(9*nr, nc/2);

 bufferTemp.realp = ( float* ) malloc (tempSize * sizeOf ( float ) );

 bufferTemp.imagp = ( float* ) malloc (tempSize * sizeOf ( float ) );

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_zrop

Computes an out-of-place single-precision real discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse).

   void vDSP_fft2d_zrop (FFTSetup setup,
   DSPSplitComplex * signal,
   vDSP_Stride signalStrideInRow,
   vDSP_Stride signalStrideInCol,
   DSPSplitComplex * result,
   vDSP_Stride strideResultInRow,
   vDSP_Stride strideResultInCol,
   vDSP_Length log2nInCol,
   vDSP_Length log2nInRow,
   FFTDirection flag);

Parameters

setup

Points to a structure initialized by a prior call to FFT weights array function vDSP_create_fftsetup or vDSP_create_fftsetupD. The value supplied as parameter log2n of the setup function must equal or exceed the value supplied as parameter log2n or log2m, whichever is larger, of the transform function.

signal

A complex vector signal input.

signalStrideInRow

Specifies a stride across each row of matrix signal. Specifying 1 for signalStrideInRow processes every element across each row, specifying 2 processes every other element across each row, and so forth.

signalStrideInCol

If not 0, represents the distance between each row of the input /output matrix. If parameter signalStrideInCol is 1024, for instance, element 512 equates to element (1,0) of matrix a, element 1024 equates to element (2,0), and so forth.

result

The complex vector signal output.

strideResultInRow

Specifies a row stride for output matrix c in the same way that signalStrideInRow specifies strides for input the matrix.

strideResultInCol

Specifies a column stride for output matrix c in the same way that signalStrideInCol specify strides for input the matrix.

log2nInCol

The base 2 exponent of the number of columns to process for each row. log2nInCol must be between 3 and 10, inclusive.

log2nInRow

flag

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of flag.

Discussion

Real data is stored in split complex form, with odd reals stored on the imaginary side of the split complex form and even reals in stored on the real side.

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_zropD

Computes an out-of-place double-precision real discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse).

   void vDSP_fft2d_zropD (FFTSetupD setup,
   DSPDoubleSplitComplex * ioData,
   vDSP_Stride Kr,
   vDSP_Stride Kc,
   DSPDoubleSplitComplex * ioData2,
   vDSP_Stride Ir,
   vDSP_Stride Ic,
   vDSP_Length log2nc,
   vDSP_Length log2nr,
   FFTDirection flag);

Parameters

setup

Points to a structure initialized by a prior call to FFT weights array function vDSP_create_fftsetup or vDSP_create_fftsetupD. The value supplied as parameter log2n of the setup function must equal or exceed the value supplied as parameter log2n or log2m, whichever is larger, of the transform function.

ioData

A complex vector input.

Kr

Specifies a stride across each row of matrix signal. Specifying 1 for Kr processes every element across each row, specifying 2 processes every other element across each row, and so forth.

Kc

If not 0, represents the distance between each row of the input /output matrix. If parameter Kc is 1024, for instance, element 512 equates to element (1,0) of matrix a, element 1024 equates to element (2,0), and so forth.

ioData2

The complex vector result.

Ir

Specifies a row stride for output matrix ioData2 in the same way that Kr specifies strides for input the matrix.

Ic

Specifies a column stride for output matrix ioData2 in the same way that Kc specify strides for input matrix ioData.

log2nc

The base 2 exponent of the number of columns to process for each row. log2nc must be between 3 and 10, inclusive.

log2nr

The base 2 exponent of the number of rows to process. For example, to process 64 rows of 128 columns, specify 7 for log2nc and 6 for log2nr. log2nr must be between 3 and 10, inclusive.

flag

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of flag.

Discussion

Real data is stored in split complex form, with odd reals stored on the imaginary side of the split complex form and even reals in stored on the real side.

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_zropt

Computes an out-of-place single-precision real discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse). A buffer is used for intermediate results.

   void vDSP_fft2d_zropt (FFTSetup setup,
   DSPSplitComplex * signal,
   vDSP_Stride signalStrideInRow,
   vDSP_Stride signalStrideInCol,
   DSPSplitComplex * result,
   vDSP_Stride strideResultInRow,
   vDSP_Stride strideResultInCol,
   DSPSplitComplex * bufferTemp,
   vDSP_Length log2nInCol,
   vDSP_Length log2nInRow,
   FFTDirection flag);

Parameters

setup

Points to a structure initialized by a prior call to FFT weights array function vDSP_create_fftsetup or vDSP_create_fftsetupD. The value supplied as parameter log2n of the setup function must equal or exceed the value supplied as parameter log2n or log2m, whichever is larger, of the transform function.

signal

A complex vector signal input.

signalStrideInRow

Specifies a stride across each row of matrix signal. Specifying 1 for signalStrideInRow processes every element across each row, specifying 2 processes every other element across each row, and so forth.

signalStrideInCol

If not 0, represents the distance between each row of matrix signal. If parameter signalStrideInCol is 1024, for instance, element 512 equates to element (1,0) of matrix signal, element 1024 equates to element (2,0), and so forth.

result

The complex vector signal output.

strideResultInRow

Specifies a row stride for output matrix result in the same way that signalStrideInRow specifies strides for input matrix result.

strideResultInCol

Specifies a column stride for output matrix c in the same way that signalStrideInCol specify strides for input matrix result.

bufferTemp

A temporary matrix used for storing interim results. The size of temporary memory for each part (real and imaginary) can be calculated using the algorithm shown below.

log2nInCol

The base 2 exponent of the number of columns to process for each row. log2nInCol must be between 3 and 10, inclusive.

log2nInRow

flag

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of flag.

Discussion

Here is the bufferTemp size algorithm:

 int nr, nc, tempSize;

 nr = 1<<log2InRow;

 nc = 1<<log2InCol;

 if ( ( (log2InCol-1) < 3 ) || ( log2InRow > 9)

 tempSize = 9 * nr;

 else

 tempSize = 17 * nr

 bufferTemp.realp = (float*) malloc (tempSize * sizeOf (float));

 bufferTemp.imagp = (float*) malloc (tempSize * sizeOf (float));

Real data is stored in split complex form, with odd reals stored on the imaginary side of the split complex form and even reals in stored on the real side.

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

vDSP_fft2d_zroptD

Computes an out-of-place double-precision real discrete FFT, either from the spatial domain to the frequency domain (forward) or from the frequency domain to the spatial domain (inverse). A buffer is used for intermediate results.

   void vDSP_fft2d_zroptD (FFTSetupD setup,
   DSPDoubleSplitComplex * ioData,
   vDSP_Stride Kr,
   vDSP_Stride Kc,
   DSPDoubleSplitComplex * ioData2,
   vDSP_Stride Ir,
   vDSP_Stride Ic,
   DSPDoubleSplitComplex * temp,
   vDSP_Length log2nc,
   vDSP_Length log2nr,
   FFTDirection flag);

Parameters

setup

Points to a structure initialized by a prior call to FFT weights array function vDSP_create_fftsetup or vDSP_create_fftsetupD. The value supplied as parameter log2n of the setup function must equal or exceed the value supplied as parameter log2n or log2m, whichever is larger, of the transform function.

ioData

A complex vector input.

Kr

Kc

ioData2

The complex vector result.

Ir

Specifies a row stride for output matrix ioData2 in the same way that Kr specifies strides for input the matrix.

Ic

Specifies a column stride for output matrix ioData2 in the same way that Kc specify strides for input matrix ioData.

temp

A temporary matrix used for storing interim results. The size of temporary memory for each part (real and imaginary) is the lower value of 16 KB or 4*n, where log2n = log2nInCol + log2nInRow.

log2nc

The base 2 exponent of the number of columns to process for each row. log2nc must be between 3 and 10, inclusive.

log2nr

The base 2 exponent of the number of rows to process. For example, to process 64 rows of 128 columns, specify 7 for log2nc and 6 for log2nr. log2nr must be between 3 and 10, inclusive.

flag

A forward/inverse directional flag, and must specify FFT_FORWARD for a forward transform or FFT_INVERSE for an inverse transform.

Results are undefined for other values of flag.

Discussion

Real data is stored in split complex form, with odd reals stored on the imaginary side of the split complex form and even reals in stored on the real side.

See also functions "vDSP_create_fftsetup", "vDSP_create_fftsetupD", "vDSP_destroy_fftsetup", and "vDSP_destroy_fftsetupD".

Availability

Available in Mac OS X v10.4 and later.

Declared In

vDSP.h

Next Page >

Hide TOC

Last updated: 2007-06-15

Did this document help you?
Yes: Tell us what works for you. It’s good, but: Report typos, inaccuracies, and so forth. It wasn’t helpful: Tell us what would have helped.




	Get information on Apple products. Visit the Apple Store online or at retail locations. 1-800-MY-APPLE Copyright © 2007 Apple Inc. All rights reserved. \| Terms of use \| Privacy Notice