Performs a wavelet based solution of the Poisson equation. More...

Functions/Subroutines
subroutine, public	psolver (geocode, iproc, nproc, n01, n02, n03, hx, hy, hz, rhopot, karray, pw_grid)
	Calculate the Poisson equation $\nabla^2 V(x,y,z)=-4 \pi \rho(x,y,z)$ from a given $\rho$, for different boundary conditions an for different data distributions. Following the boundary conditions, it applies the Poisson Kernel previously calculated. More...

subroutine, public	p_fft_dimensions (n01, n02, n03, m1, m2, m3, n1, n2, n3, md1, md2, md3, nd1, nd2, nd3, nproc)
	Calculate four sets of dimension needed for the calculation of the convolution for the periodic system. More...

subroutine, public	s_fft_dimensions (n01, n02, n03, m1, m2, m3, n1, n2, n3, md1, md2, md3, nd1, nd2, nd3, nproc)
	Calculate four sets of dimension needed for the calculation of the convolution for the surface system. More...

subroutine, public	f_fft_dimensions (n01, n02, n03, m1, m2, m3, n1, n2, n3, md1, md2, md3, nd1, nd2, nd3, nproc)
	Calculate four sets of dimension needed for the calculation of the zero-padded convolution. More...

Detailed Description

Performs a wavelet based solution of the Poisson equation.

Author: Florian Schiffmann (09.2007,fschiff)

Function/Subroutine Documentation

◆ psolver()

subroutine, public ps_wavelet_util::psolver	(	character(len=1), intent(in)	geocode,
		integer, intent(in)	iproc,
		integer, intent(in)	nproc,
		integer, intent(in)	n01,
		integer, intent(in)	n02,
		integer, intent(in)	n03,
		real(kind=dp), intent(in)	hx,
		real(kind=dp), intent(in)	hy,
		real(kind=dp), intent(in)	hz,
		real(kind=dp), dimension(*), intent(inout)	rhopot,
		real(kind=dp), dimension(*), intent(in)	karray,
		type(pw_grid_type), pointer	pw_grid
	)

Calculate the Poisson equation $\nabla^2 V(x,y,z)=-4 \pi \rho(x,y,z)$ from a given $\rho$, for different boundary conditions an for different data distributions. Following the boundary conditions, it applies the Poisson Kernel previously calculated.

Parameters

geocode	Indicates the boundary conditions (BC) of the problem: 'F' free BC, isolated systems. The program calculates the solution as if the given density is "alone" in R^3 space. 'S' surface BC, isolated in y direction, periodic in xz plane The given density is supposed to be periodic in the xz plane, so the dimensions in these direction mus be compatible with the FFT Beware of the fact that the isolated direction is y! 'P' periodic BC. The density is supposed to be periodic in all the three directions, then all the dimensions must be compatible with the FFT. No need for setting up the kernel.
iproc	label of the process,from 0 to nproc-1
nproc	number of processors
n01	global dimension in the three directions.
n02	global dimension in the three directions.
n03	global dimension in the three directions.
hx	grid spacings. For the isolated BC case for the moment they are supposed to be equal in the three directions
hy	grid spacings. For the isolated BC case for the moment they are supposed to be equal in the three directions
hz	grid spacings. For the isolated BC case for the moment they are supposed to be equal in the three directions
rhopot	main input/output array. On input, it represents the density values on the grid points On output, it is the Hartree potential, namely the solution of the Poisson equation PLUS (when ixc/=0) the XC potential PLUS (again for ixc/=0) the pot_ion array. The output is non overlapping, in the sense that it does not consider the points that are related to gradient and WB calculation
karray	kernel of the poisson equation. It is provided in distributed case, with dimensions that are related to the output of the PS_dim4allocation routine it MUST be created by following the same geocode as the Poisson Solver.
pw_grid	...

Date: February 2007

Author: Luigi Genovese

Note: The dimensions of the arrays must be compatible with geocode, nproc, ixc and iproc. Since the arguments of these routines are indicated with the *, it is IMPERATIVE to use the PS_dim4allocation routine for calculation arrays sizes.

Definition at line 81 of file ps_wavelet_util.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ p_fft_dimensions()

subroutine, public ps_wavelet_util::p_fft_dimensions	(	integer, intent(in)	n01,
		integer, intent(in)	n02,
		integer, intent(in)	n03,
		integer, intent(out)	m1,
		integer, intent(out)	m2,
		integer, intent(out)	m3,
		integer, intent(out)	n1,
		integer, intent(out)	n2,
		integer, intent(out)	n3,
		integer, intent(out)	md1,
		integer, intent(out)	md2,
		integer, intent(out)	md3,
		integer, intent(out)	nd1,
		integer, intent(out)	nd2,
		integer, intent(out)	nd3,
		integer, intent(in)	nproc
	)

Calculate four sets of dimension needed for the calculation of the convolution for the periodic system.

Parameters

n01	original real dimensions (input)
n02	original real dimensions (input)
n03	original real dimensions (input)
m1	original real dimension, with m2 and m3 exchanged
m2	original real dimension, with m2 and m3 exchanged
m3	original real dimension, with m2 and m3 exchanged
n1	the first FFT dimensions, for the moment supposed to be even
n2	the first FFT dimensions, for the moment supposed to be even
n3	the first FFT dimensions, for the moment supposed to be even
md1	the n1,n2,n3 dimensions. They contain the real unpadded space, properly enlarged to be compatible with the FFT dimensions n_i. md2 is further enlarged to be a multiple of nproc
md2	the n1,n2,n3 dimensions. They contain the real unpadded space, properly enlarged to be compatible with the FFT dimensions n_i. md2 is further enlarged to be a multiple of nproc
md3	the n1,n2,n3 dimensions. They contain the real unpadded space, properly enlarged to be compatible with the FFT dimensions n_i. md2 is further enlarged to be a multiple of nproc
nd1	fourier dimensions for which the kernel is injective, formally 1/8 of the fourier grid. Here the dimension nd3 is enlarged to be a multiple of nproc
nd2	fourier dimensions for which the kernel is injective, formally 1/8 of the fourier grid. Here the dimension nd3 is enlarged to be a multiple of nproc
nd3	fourier dimensions for which the kernel is injective, formally 1/8 of the fourier grid. Here the dimension nd3 is enlarged to be a multiple of nproc
nproc	...

Date: October 2006

Author: Luigi Genovese

Note: This four sets of dimensions are actually redundant (mi=n0i), due to the backward-compatibility with the other geometries of the Poisson Solver. The dimensions 2 and 3 are exchanged.

Definition at line 237 of file ps_wavelet_util.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ s_fft_dimensions()

subroutine, public ps_wavelet_util::s_fft_dimensions	(	integer, intent(in)	n01,
		integer, intent(in)	n02,
		integer, intent(in)	n03,
		integer, intent(out)	m1,
		integer, intent(out)	m2,
		integer, intent(out)	m3,
		integer, intent(out)	n1,
		integer, intent(out)	n2,
		integer, intent(out)	n3,
		integer, intent(out)	md1,
		integer, intent(out)	md2,
		integer, intent(out)	md3,
		integer, intent(out)	nd1,
		integer, intent(out)	nd2,
		integer, intent(out)	nd3,
		integer, intent(in)	nproc
	)

Calculate four sets of dimension needed for the calculation of the convolution for the surface system.

Parameters

n01	original real dimensions (input)
n02	original real dimensions (input)
n03	original real dimensions (input)
m1	original real dimension, with 2 and 3 exchanged
m2	original real dimension, with 2 and 3 exchanged
m3	original real dimension, with 2 and 3 exchanged
n1	the first FFT dimensions, for the moment supposed to be even
n2	the first FFT dimensions, for the moment supposed to be even
n3	the double of the first FFT even dimension greater than m3 (improved for the HalFFT procedure)
md1	the n1,n2 dimensions.
md2	the n1,n2,n3 dimensions.
md3	the half of n3 dimension. They contain the real unpadded space, properly enlarged to be compatible with the FFT dimensions n_i. md2 is further enlarged to be a multiple of nproc
nd1	fourier dimensions for which the kernel is injective, formally 1/8 of the fourier grid. Here the dimension nd3 is enlarged to be a multiple of nproc
nd2	fourier dimensions for which the kernel is injective, formally 1/8 of the fourier grid. Here the dimension nd3 is enlarged to be a multiple of nproc
nd3	fourier dimensions for which the kernel is injective, formally 1/8 of the fourier grid. Here the dimension nd3 is enlarged to be a multiple of nproc
nproc	...

Date: October 2006

Author: Luigi Genovese

Note: This four sets of dimensions are actually redundant (mi=n0i), due to the backward-compatibility with the Poisson Solver with other geometries. Dimensions n02 and n03 were exchanged

Definition at line 333 of file ps_wavelet_util.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ f_fft_dimensions()

subroutine, public ps_wavelet_util::f_fft_dimensions	(	integer, intent(in)	n01,
		integer, intent(in)	n02,
		integer, intent(in)	n03,
		integer, intent(out)	m1,
		integer, intent(out)	m2,
		integer, intent(out)	m3,
		integer, intent(out)	n1,
		integer, intent(out)	n2,
		integer, intent(out)	n3,
		integer, intent(out)	md1,
		integer, intent(out)	md2,
		integer, intent(out)	md3,
		integer, intent(out)	nd1,
		integer, intent(out)	nd2,
		integer, intent(out)	nd3,
		integer, intent(in)	nproc
	)

Calculate four sets of dimension needed for the calculation of the zero-padded convolution.

Parameters

n01	original real dimensions (input)
n02	original real dimensions (input)
n03	original real dimensions (input)
m1	original real dimension with the dimension 2 and 3 exchanged
m2	original real dimension with the dimension 2 and 3 exchanged
m3	original real dimension with the dimension 2 and 3 exchanged
n1	...
n2	...
n3	the double of the first FFT even dimension greater than m3 (improved for the HalFFT procedure)
md1	half of n1,n2,n3 dimension. They contain the real unpadded space, properly enlarged to be compatible with the FFT dimensions n_i. md2 is further enlarged to be a multiple of nproc
md2	half of n1,n2,n3 dimension. They contain the real unpadded space, properly enlarged to be compatible with the FFT dimensions n_i. md2 is further enlarged to be a multiple of nproc
md3	half of n1,n2,n3 dimension. They contain the real unpadded space, properly enlarged to be compatible with the FFT dimensions n_i. md2 is further enlarged to be a multiple of nproc
nd1	fourier dimensions for which the kernel FFT is injective, formally 1/8 of the fourier grid. Here the dimension nd3 is enlarged to be a multiple of nproc
nd2	fourier dimensions for which the kernel FFT is injective, formally 1/8 of the fourier grid. Here the dimension nd3 is enlarged to be a multiple of nproc
nd3	fourier dimensions for which the kernel FFT is injective, formally 1/8 of the fourier grid. Here the dimension nd3 is enlarged to be a multiple of nproc
nproc	...

Date: February 2006

Author: Luigi Genovese

Note: The dimension m2 and m3 correspond to n03 and n02 respectively this is needed since the convolution routine manage arrays of dimension (md1,md3,md2/nproc)

Definition at line 440 of file ps_wavelet_util.F.

Here is the call graph for this function:

Here is the caller graph for this function:

Functions/Subroutines

Detailed Description

Function/Subroutine Documentation

◆ psolver()

◆ p_fft_dimensions()

◆ s_fft_dimensions()

◆ f_fft_dimensions()