Fourier transform wrapper for real or complex data. More...

#include <FFT.h>

Public Member Functions
	FFT ()
	Default constructor.
virtual	~FFT ()
	Destructor.
void	setup (IntVec< D > const &meshDimensions)
	Setup grid dimensions, plans and work space.
void	forwardTransform (RField< D > const &rField, RFieldDft< D > &kField) const
	Compute forward (real-to-complex) discrete Fourier transform.
void	inverseTransformUnsafe (RFieldDft< D > &kField, RField< D > &rField) const
	Compute inverse (complex-to-real) DFT, overwriting the input.
void	inverseTransformSafe (RFieldDft< D > const &kField, RField< D > &rField) const
	Compute inverse (complex-to-real) DFT without overwriting input.
void	forwardTransform (CField< D > const &rField, CField< D > &kField) const
	Compute forward (complex-to-complex) discrete Fourier transform.
void	inverseTransform (CField< D > const &kField, CField< D > &rField) const
	Compute inverse (complex-to-complex) discrete Fourier transform.
const IntVec< D > &	meshDimensions () const
	Return the dimensions of the grid for which this was allocated.
bool	isSetup () const
	Has this FFT object been setup?

Static Public Member Functions
static void	computeKMesh (IntVec< D > const &rMeshDimensions, IntVec< D > &kMeshDimensions, int &kSize)
	Compute dimensions and size of k-space mesh for DFT of real data.
static bool	hasImplicitInverse (IntVec< D > const &wavevector, IntVec< D > const &meshDimensions)
	Does a wavevector have an implicit inverse in DFT for real data?

Detailed Description

template<int D>
class Pscf::Prdc::Cuda::FFT< D >

Fourier transform wrapper for real or complex data.

This class is a wrapper for plan creation and discrete Fourier transform (DFT) functions provided by the NVIDIA cufft library, providing an interface to the field container classes RField<D>, RField<Dft>, and CField<D> in namespace Pscf::Prdc::Cuda.

Definition at line 38 of file cuda/FFT.h.

Constructor & Destructor Documentation

◆ FFT()

template<int D>

Pscf::Prdc::Cuda::FFT< D >::FFT ( )

Default constructor.

◆ ~FFT()

template<int D>

virtual Pscf::Prdc::Cuda::FFT< D >::~FFT ( )

virtual

Destructor.

Member Function Documentation

◆ setup()

template<int D>

void Pscf::Prdc::Cuda::FFT< D >::setup ( IntVec< D > const & meshDimensions )

Setup grid dimensions, plans and work space.

Parameters

meshDimensions Dimensions of real-space grid.

References meshDimensions().

◆ forwardTransform() [1/2]

template<int D>

void Pscf::Prdc::Cuda::FFT< D >::forwardTransform	(	RField< D > const &	rField,
		RFieldDft< D > &	kField ) const

Compute forward (real-to-complex) discrete Fourier transform.

The resulting complex array is the output of cuFFT's forward transform method scaled by a factor of 1/N (where N is the number of grid points). The scaling ensures that a round-trip Fourier transform (forward + inverse) reproduces the original input array.

This transform is inherently "safe", meaning that the input array will not be modified during the DFT.

Parameters

rField	real values on r-space grid (input, gpu mem)
kField	complex values on k-space grid (output, gpu mem)

◆ inverseTransformUnsafe()

template<int D>

void Pscf::Prdc::Cuda::FFT< D >::inverseTransformUnsafe	(	RFieldDft< D > &	kField,
		RField< D > &	rField ) const

Compute inverse (complex-to-real) DFT, overwriting the input.

The resulting real array is the unaltered output of cuFFT's inverse transform function.

This transform is inherently "unsafe", meaning that the input array will be overwritten during the DFT. Accordingly, the input array kField is not const like it is in all the other transform methods.

Parameters

kField	complex values on k-space grid (input, gpu mem)
rField	real values on r-space grid (output, gpu mem)

◆ inverseTransformSafe()

template<int D>

void Pscf::Prdc::Cuda::FFT< D >::inverseTransformSafe	(	RFieldDft< D > const &	kField,
		RField< D > &	rField ) const

Compute inverse (complex-to-real) DFT without overwriting input.

This method makes a copy of the input kField array before performing the DFT, thus preserving the input.

Parameters

kField	complex values on k-space grid (input, gpu mem)
rField	real values on r-space grid (output, gpu mem)

◆ forwardTransform() [2/2]

template<int D>

void Pscf::Prdc::Cuda::FFT< D >::forwardTransform	(	CField< D > const &	rField,
		CField< D > &	kField ) const

Compute forward (complex-to-complex) discrete Fourier transform.

The resulting complex array is the output of cuFFT's forward transform method scaled by a factor of 1/N (where N is the number of grid points). The scaling ensures that a round-trip Fourier transform (forward + inverse) reproduces the original input array.

This transform is inherently "safe", meaning that the input array will not be modified during the DFT.

Note that, in this transform, the k-space grid is the same size as the r-space grid, which differs from the real-to-complex transform.

Parameters

rField	complex values on r-space grid (input, gpu mem)
kField	complex values on k-space grid (output, gpu mem)

◆ inverseTransform()

template<int D>

void Pscf::Prdc::Cuda::FFT< D >::inverseTransform	(	CField< D > const &	kField,
		CField< D > &	rField ) const

Compute inverse (complex-to-complex) discrete Fourier transform.

The resulting complex array is the unaltered output of cuFFT's inverse transform function.

This transform is inherently "safe", meaning that the input array will not be modified during the DFT.

Note that, in this transform, the k-space grid is the same size as the r-space grid, which differs from the real-to-complex transform.

Parameters

kField	complex values on k-space grid (input, gpu mem)
rField	complex values on r-space grid (output, gpu mem)

◆ meshDimensions()

template<int D>

const IntVec< D > & Pscf::Prdc::Cuda::FFT< D >::meshDimensions ( ) const

Return the dimensions of the grid for which this was allocated.

Referenced by hasImplicitInverse(), and setup().

◆ isSetup()

template<int D>

bool Pscf::Prdc::Cuda::FFT< D >::isSetup ( ) const

Has this FFT object been setup?

◆ computeKMesh()

template<int D>

void Pscf::Prdc::Cuda::FFT< D >::computeKMesh	(	IntVec< D > const &	rMeshDimensions,
		IntVec< D > &	kMeshDimensions,
		int &	kSize )

static

Compute dimensions and size of k-space mesh for DFT of real data.

A corresponding function is not needed for complex-to-complex transforms because the real-space and Fourier-space grids have the same dimensions in this case.

Parameters

rMeshDimensions	dimensions of real space grid (real data)
kMeshDimensions	dimensions of k-space grid (complex data)
kSize	number of point in k-space grid

Referenced by Pscf::Prdc::Cuda::RFieldDft< D >::associate().

◆ hasImplicitInverse()

template<int D>

bool Pscf::Prdc::Cuda::FFT< D >::hasImplicitInverse	(	IntVec< D > const &	wavevector,
		IntVec< D > const &	meshDimensions )

static

Does a wavevector have an implicit inverse in DFT for real data?

The inverse of a wavevector within the DFT mesh used for a DFT of real data is "implicit" if its inverse is not equivalent to any wavevector in this k-mesh.

To compute a Fourier sum using the DFT of real r-space data, for any summand that has inversion symmetry (i.e., the same value for every vector and its inverse), sum over all vectors in the k-space mesh and multiply each element for each wavevector that has an implicit inverse by 2.0, and all others by 1.0.

Parameters

wavevector	integer indices of wavevector (in)
meshDimensions	dimensions of real-space mesh (in)

Returns: true iff inverse is implicit (i.e., not in DFT k-mesh)

References meshDimensions().

The documentation for this class was generated from the following file:

cuda/FFT.h

Public Member Functions

Static Public Member Functions

Detailed Description

Constructor & Destructor Documentation

◆ FFT()

◆ ~FFT()

Member Function Documentation

◆ setup()

◆ forwardTransform() [1/2]

◆ inverseTransformUnsafe()

◆ inverseTransformSafe()

◆ forwardTransform() [2/2]

◆ inverseTransform()

◆ meshDimensions()

◆ isSetup()

◆ computeKMesh()

◆ hasImplicitInverse()