Collaboration diagram for Fast Gauss transfrom with complex parameter:

Data Structures
struct	fgt_plan
	Structure for the Gauss transform. More...

Macros
#define	NFFT_PRECISION_DOUBLE

#define	DGT_PRE_CEXP (1U<< 0)
	If this flag is set, the whole matrix is precomputed and stored for the discrete Gauss transfrom.

#define	FGT_NDFT (1U<< 1)
	If this flag is set, the fast Gauss transform uses the discrete instead of the fast Fourier transform.

#define	FGT_APPROX_B (1U<< 2)
	If this flag is set, the discrete Fourier coefficients of the uniformly sampled Gaussian are used instead of the sampled continuous Fourier transform.

Functions
static void	dgt_trafo (fgt_plan *ths)
	Executes the discrete Gauss transform.

static void	fgt_trafo (fgt_plan *ths)
	Executes the fast Gauss transform.

static void	fgt_init_guru (fgt_plan *ths, int N, int M, NFFT_C sigma, int n, NFFT_R p, int m, unsigned flags)
	Initialisation of a transform plan, guru.

static void	fgt_init (fgt_plan *ths, int N, int M, NFFT_C sigma, NFFT_R eps)
	Initialisation of a transform plan, simple.

static void	fgt_init_node_dependent (fgt_plan *ths)
	Initialisation of a transform plan, depends on source and target nodes.

static void	fgt_finalize (fgt_plan *ths)
	Destroys the transform plan.

static void	fgt_test_init_rand (fgt_plan *ths)
	Random initialisation of a fgt plan.

static NFFT_R	fgt_test_measure_time (fgt_plan *ths, unsigned dgt)
	Compares execution times for the fast and discrete Gauss transform.

static void	fgt_test_simple (int N, int M, NFFT_C sigma, NFFT_R eps)
	Simple example that computes fast and discrete Gauss transforms.

static void	fgt_test_andersson (void)
	Compares accuracy and execution time of the fast Gauss transform with increasing expansion degree.

static void	fgt_test_error (void)
	Compares accuracy of the fast Gauss transform with increasing expansion degree.

static void	fgt_test_error_p (void)
	Compares accuracy of the fast Gauss transform with increasing expansion degree and different periodisation lengths.

int	main (int argc, char **argv)
	Different tests of the fast Gauss transform.

Variables
int	fgt_plan::N
	number of source nodes

int	fgt_plan::M
	number of target nodes

NFFT_C *	fgt_plan::alpha
	source coefficients

NFFT_C *	fgt_plan::f
	target evaluations

unsigned	fgt_plan::flags
	flags for precomputation and approximation type

NFFT_C	fgt_plan::sigma
	parameter of the Gaussian

NFFT_R *	fgt_plan::x
	source nodes in

NFFT_R *	fgt_plan::y
	target nodes in

NFFT_C *	fgt_plan::pre_cexp
	precomputed values for dgt

int	fgt_plan::n
	expansion degree

NFFT_R	fgt_plan::p
	period, at least 1

NFFT_C *	fgt_plan::b
	expansion coefficients

Detailed Description

Macro Definition Documentation

◆ NFFT_PRECISION_DOUBLE

#define NFFT_PRECISION_DOUBLE

Definition at line 30 of file fastgauss.c.

◆ DGT_PRE_CEXP

#define DGT_PRE_CEXP (1U<< 0)

If this flag is set, the whole matrix is precomputed and stored for the discrete Gauss transfrom.

See also: fgt_init_node_dependent; fgt_init

Author: Stefan Kunis

Definition at line 42 of file fastgauss.c.

◆ FGT_NDFT

#define FGT_NDFT (1U<< 1)

If this flag is set, the fast Gauss transform uses the discrete instead of the fast Fourier transform.

See also: fgt_init; nfft_trafo_direct; nfft_trafo

Author: Stefan Kunis

Definition at line 53 of file fastgauss.c.

◆ FGT_APPROX_B

#define FGT_APPROX_B (1U<< 2)

If this flag is set, the discrete Fourier coefficients of the uniformly sampled Gaussian are used instead of the sampled continuous Fourier transform.