2016-02-14 20:49:16 +08:00
|
|
|
!
|
|
|
|
! Copyright (C) 2001-2016 Quantum ESPRESSO group
|
|
|
|
! This file is distributed under the terms of the
|
|
|
|
! GNU General Public License. See the file `License'
|
|
|
|
! in the root directory of the present distribution,
|
|
|
|
! or http://www.gnu.org/copyleft/gpl.txt .
|
|
|
|
!
|
|
|
|
!-----------------------------------------------------------------------
|
2020-02-29 15:59:35 +08:00
|
|
|
subroutine incdrhoscf_nc (drhoscf, weight, ik, dbecsum, dpsi, rsign)
|
2016-02-14 20:49:16 +08:00
|
|
|
!-----------------------------------------------------------------------
|
|
|
|
!
|
|
|
|
! This routine computes the change of the charge density due to the
|
|
|
|
! perturbation. It is called at the end of the computation of the
|
|
|
|
! change of the wavefunction for a given k point.
|
|
|
|
!
|
|
|
|
USE kinds, ONLY : DP
|
|
|
|
USE ions_base, ONLY : nat
|
|
|
|
USE cell_base, ONLY : omega
|
MAJOR restructuring of the FFTXlib library
In real space processors are organized in a 2D pattern.
Each processor owns data from a sub-set of Z-planes and a sub-set of Y-planes.
In reciprocal space each processor owns Z-columns that belong to a sub set of
X-values. This allows to split the processors in two sets for communication
in the YZ and XY planes.
In alternative, if the situation allows for it, a task group paralelization is used
(with ntg=nyfft) where complete XY planes of ntg wavefunctions are collected and Fourier
trasnformed in G space by different task-groups. This is preferable to the Z-proc + Y-proc
paralleization if task group can be used because a smaller number of larger ammounts of
data are transferred. Hence three types of fft are implemented:
!
!! ... isgn = +-1 : parallel 3d fft for rho and for the potential
!
!! ... isgn = +-2 : parallel 3d fft for wavefunctions
!
!! ... isgn = +-3 : parallel 3d fft for wavefunctions with task group
!
!! ... isgn = + : G-space to R-space, output = \sum_G f(G)exp(+iG*R)
!! ... fft along z using pencils (cft_1z)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along x using pencils (cft_1x)
!
!! ... isgn = - : R-space to G-space, output = \int_R f(R)exp(-iG*R)/Omega
!! ... fft along x using pencils (cft_1x)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along z using pencils (cft_1z)
!
! If task_group_fft_is_active the FFT acts on a number of wfcs equal to
! dfft%nproc2, the number of Y-sections in which a plane is divided.
! Data are reshuffled by the fft_scatter_tg routine so that each of the
! dfft%nproc2 subgroups (made by dfft%nproc3 procs) deals with whole planes
! of a single wavefunciton.
!
fft_type module heavily modified, a number of variables renamed with more intuitive names
(at least to me), a number of more variables introduced for the Y-proc parallelization.
Task_group module made void. task_group management is now reduced to the logical component
fft_desc%have_task_groups of fft_type_descriptor type variable fft_desc.
In term of interfaces, the 'easy' calling sequences are
SUBROUTINE invfft/fwfft( grid_type, f, dfft, howmany )
!! where:
!!
!! **grid_type = 'Dense'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the dense grid (dfftp). On output, f is overwritten
!!
!! **grid_type = 'Smooth'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Wave'** :
!! inverse/direct fourier transform of wave functions f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'tgWave'** :
!! inverse/direct fourier transform of wave functions f with task group
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Custom'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **grid_type = 'CustomWave'** :
!! inverse/direct fourier transform of wave functions f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **dfft = FFT descriptor**, IMPORTANT NOTICE: grid is specified only by dfft.
!! No check is performed on the correspondence between dfft and grid_type.
!! grid_type is now used only to distinguish cases 'Wave' / 'CustomWave'
!! from all other cases
Many more files modified.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@13676 c92efa57-630b-4861-b058-cf58834340f0
2017-08-02 04:31:02 +08:00
|
|
|
USE fft_base, ONLY : dffts, dfftp
|
2016-02-14 20:49:16 +08:00
|
|
|
USE fft_interfaces, ONLY : invfft
|
|
|
|
USE lsda_mod, ONLY : nspin
|
|
|
|
USE spin_orb, ONLY : domag
|
|
|
|
USE noncollin_module, ONLY : npol, nspin_mag
|
|
|
|
USE uspp_param, ONLY : nhm
|
2016-06-05 04:18:02 +08:00
|
|
|
USE wvfct, ONLY : npwx, nbnd
|
2018-07-12 22:02:58 +08:00
|
|
|
USE wavefunctions, ONLY : evc
|
2016-06-05 04:18:02 +08:00
|
|
|
USE klist, ONLY : ngk,igk_k
|
|
|
|
USE qpoint, ONLY : ikks, ikqs
|
2016-02-14 20:49:16 +08:00
|
|
|
USE control_lr, ONLY : nbnd_occ
|
2020-01-29 18:44:02 +08:00
|
|
|
USE qpoint_aux, ONLY : becpt
|
|
|
|
USE lrus, ONLY : becp1
|
2016-02-14 20:49:16 +08:00
|
|
|
USE mp_bands, ONLY : me_bgrp, inter_bgrp_comm, ntask_groups
|
|
|
|
USE mp, ONLY : mp_sum
|
2017-08-27 00:54:46 +08:00
|
|
|
USE fft_helper_subroutines
|
2016-02-14 20:49:16 +08:00
|
|
|
|
|
|
|
IMPLICIT NONE
|
|
|
|
!
|
|
|
|
! I/O variables
|
|
|
|
INTEGER, INTENT(IN) :: ik
|
|
|
|
! input: the k point
|
|
|
|
REAL(DP), INTENT(IN) :: weight
|
2020-02-29 15:59:35 +08:00
|
|
|
REAL(DP), INTENT(IN) :: rsign
|
2016-02-14 20:49:16 +08:00
|
|
|
! input: the weight of the k point
|
2020-02-29 15:59:35 +08:00
|
|
|
! the sign in front of the response of the magnetization density
|
2016-02-14 20:49:16 +08:00
|
|
|
COMPLEX(DP), INTENT(IN) :: dpsi(npwx*npol,nbnd)
|
|
|
|
! input: the perturbed wfcs at the given k point
|
|
|
|
COMPLEX(DP), INTENT(INOUT) :: drhoscf (dfftp%nnr,nspin_mag), dbecsum (nhm,nhm,nat,nspin)
|
|
|
|
! input/output: the accumulated change of the charge density and dbecsum
|
|
|
|
!
|
|
|
|
! here the local variable
|
|
|
|
!
|
|
|
|
REAL(DP) :: wgt
|
|
|
|
! the effective weight of the k point
|
|
|
|
!
|
|
|
|
COMPLEX(DP), ALLOCATABLE :: psi (:,:), dpsic (:,:)
|
|
|
|
! the wavefunctions in real space
|
|
|
|
! the change of wavefunctions in real space
|
|
|
|
!
|
|
|
|
COMPLEX(DP), ALLOCATABLE :: tg_psi (:,:), tg_dpsi (:,:), tg_drho(:,:)
|
|
|
|
!
|
2016-06-05 04:18:02 +08:00
|
|
|
INTEGER :: npw, npwq, ikk, ikq
|
MAJOR restructuring of the FFTXlib library
In real space processors are organized in a 2D pattern.
Each processor owns data from a sub-set of Z-planes and a sub-set of Y-planes.
In reciprocal space each processor owns Z-columns that belong to a sub set of
X-values. This allows to split the processors in two sets for communication
in the YZ and XY planes.
In alternative, if the situation allows for it, a task group paralelization is used
(with ntg=nyfft) where complete XY planes of ntg wavefunctions are collected and Fourier
trasnformed in G space by different task-groups. This is preferable to the Z-proc + Y-proc
paralleization if task group can be used because a smaller number of larger ammounts of
data are transferred. Hence three types of fft are implemented:
!
!! ... isgn = +-1 : parallel 3d fft for rho and for the potential
!
!! ... isgn = +-2 : parallel 3d fft for wavefunctions
!
!! ... isgn = +-3 : parallel 3d fft for wavefunctions with task group
!
!! ... isgn = + : G-space to R-space, output = \sum_G f(G)exp(+iG*R)
!! ... fft along z using pencils (cft_1z)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along x using pencils (cft_1x)
!
!! ... isgn = - : R-space to G-space, output = \int_R f(R)exp(-iG*R)/Omega
!! ... fft along x using pencils (cft_1x)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along z using pencils (cft_1z)
!
! If task_group_fft_is_active the FFT acts on a number of wfcs equal to
! dfft%nproc2, the number of Y-sections in which a plane is divided.
! Data are reshuffled by the fft_scatter_tg routine so that each of the
! dfft%nproc2 subgroups (made by dfft%nproc3 procs) deals with whole planes
! of a single wavefunciton.
!
fft_type module heavily modified, a number of variables renamed with more intuitive names
(at least to me), a number of more variables introduced for the Y-proc parallelization.
Task_group module made void. task_group management is now reduced to the logical component
fft_desc%have_task_groups of fft_type_descriptor type variable fft_desc.
In term of interfaces, the 'easy' calling sequences are
SUBROUTINE invfft/fwfft( grid_type, f, dfft, howmany )
!! where:
!!
!! **grid_type = 'Dense'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the dense grid (dfftp). On output, f is overwritten
!!
!! **grid_type = 'Smooth'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Wave'** :
!! inverse/direct fourier transform of wave functions f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'tgWave'** :
!! inverse/direct fourier transform of wave functions f with task group
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Custom'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **grid_type = 'CustomWave'** :
!! inverse/direct fourier transform of wave functions f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **dfft = FFT descriptor**, IMPORTANT NOTICE: grid is specified only by dfft.
!! No check is performed on the correspondence between dfft and grid_type.
!! grid_type is now used only to distinguish cases 'Wave' / 'CustomWave'
!! from all other cases
Many more files modified.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@13676 c92efa57-630b-4861-b058-cf58834340f0
2017-08-02 04:31:02 +08:00
|
|
|
INTEGER :: ibnd, jbnd, ir, ir3, ig, incr, v_siz, idx, ioff, ioff_tg, nxyp
|
2017-08-27 00:54:46 +08:00
|
|
|
INTEGER :: ntgrp, right_inc
|
2016-02-14 20:49:16 +08:00
|
|
|
! counters
|
|
|
|
!
|
2020-01-29 18:44:02 +08:00
|
|
|
!
|
2016-02-14 20:49:16 +08:00
|
|
|
CALL start_clock ('incdrhoscf')
|
|
|
|
!
|
|
|
|
ALLOCATE (dpsic(dffts%nnr, npol))
|
|
|
|
ALLOCATE (psi (dffts%nnr, npol))
|
|
|
|
!
|
|
|
|
wgt = 2.d0 * weight / omega
|
|
|
|
ikk = ikks(ik)
|
2016-06-05 04:18:02 +08:00
|
|
|
ikq = ikqs(ik)
|
|
|
|
npw = ngk(ikk)
|
|
|
|
npwq= ngk(ikq)
|
2016-02-14 20:49:16 +08:00
|
|
|
incr = 1
|
|
|
|
!
|
2018-01-03 08:38:48 +08:00
|
|
|
IF (dffts%has_task_groups) THEN
|
2016-02-14 20:49:16 +08:00
|
|
|
!
|
MAJOR restructuring of the FFTXlib library
In real space processors are organized in a 2D pattern.
Each processor owns data from a sub-set of Z-planes and a sub-set of Y-planes.
In reciprocal space each processor owns Z-columns that belong to a sub set of
X-values. This allows to split the processors in two sets for communication
in the YZ and XY planes.
In alternative, if the situation allows for it, a task group paralelization is used
(with ntg=nyfft) where complete XY planes of ntg wavefunctions are collected and Fourier
trasnformed in G space by different task-groups. This is preferable to the Z-proc + Y-proc
paralleization if task group can be used because a smaller number of larger ammounts of
data are transferred. Hence three types of fft are implemented:
!
!! ... isgn = +-1 : parallel 3d fft for rho and for the potential
!
!! ... isgn = +-2 : parallel 3d fft for wavefunctions
!
!! ... isgn = +-3 : parallel 3d fft for wavefunctions with task group
!
!! ... isgn = + : G-space to R-space, output = \sum_G f(G)exp(+iG*R)
!! ... fft along z using pencils (cft_1z)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along x using pencils (cft_1x)
!
!! ... isgn = - : R-space to G-space, output = \int_R f(R)exp(-iG*R)/Omega
!! ... fft along x using pencils (cft_1x)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along z using pencils (cft_1z)
!
! If task_group_fft_is_active the FFT acts on a number of wfcs equal to
! dfft%nproc2, the number of Y-sections in which a plane is divided.
! Data are reshuffled by the fft_scatter_tg routine so that each of the
! dfft%nproc2 subgroups (made by dfft%nproc3 procs) deals with whole planes
! of a single wavefunciton.
!
fft_type module heavily modified, a number of variables renamed with more intuitive names
(at least to me), a number of more variables introduced for the Y-proc parallelization.
Task_group module made void. task_group management is now reduced to the logical component
fft_desc%have_task_groups of fft_type_descriptor type variable fft_desc.
In term of interfaces, the 'easy' calling sequences are
SUBROUTINE invfft/fwfft( grid_type, f, dfft, howmany )
!! where:
!!
!! **grid_type = 'Dense'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the dense grid (dfftp). On output, f is overwritten
!!
!! **grid_type = 'Smooth'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Wave'** :
!! inverse/direct fourier transform of wave functions f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'tgWave'** :
!! inverse/direct fourier transform of wave functions f with task group
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Custom'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **grid_type = 'CustomWave'** :
!! inverse/direct fourier transform of wave functions f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **dfft = FFT descriptor**, IMPORTANT NOTICE: grid is specified only by dfft.
!! No check is performed on the correspondence between dfft and grid_type.
!! grid_type is now used only to distinguish cases 'Wave' / 'CustomWave'
!! from all other cases
Many more files modified.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@13676 c92efa57-630b-4861-b058-cf58834340f0
2017-08-02 04:31:02 +08:00
|
|
|
v_siz = dffts%nnr_tg
|
2016-02-14 20:49:16 +08:00
|
|
|
!
|
|
|
|
ALLOCATE( tg_psi( v_siz, npol ) )
|
|
|
|
ALLOCATE( tg_dpsi( v_siz, npol ) )
|
|
|
|
ALLOCATE( tg_drho( v_siz, nspin_mag ) )
|
|
|
|
!
|
2017-08-27 00:54:46 +08:00
|
|
|
incr = fftx_ntgrp(dffts)
|
2016-02-14 20:49:16 +08:00
|
|
|
!
|
|
|
|
ENDIF
|
|
|
|
!
|
|
|
|
! dpsi contains the perturbed wavefunctions of this k point
|
|
|
|
! evc contains the unperturbed wavefunctions of this k point
|
|
|
|
!
|
|
|
|
do ibnd = 1, nbnd_occ(ikk), incr
|
|
|
|
|
2018-01-03 08:38:48 +08:00
|
|
|
IF (dffts%has_task_groups) THEN
|
2016-02-14 20:49:16 +08:00
|
|
|
!
|
|
|
|
tg_drho=(0.0_DP, 0.0_DP)
|
|
|
|
tg_psi=(0.0_DP, 0.0_DP)
|
|
|
|
tg_dpsi=(0.0_DP, 0.0_DP)
|
|
|
|
!
|
|
|
|
ioff = 0
|
2017-08-27 00:54:46 +08:00
|
|
|
CALL tg_get_recip_inc( dffts, right_inc )
|
|
|
|
ntgrp = fftx_ntgrp( dffts )
|
2016-02-14 20:49:16 +08:00
|
|
|
!
|
2017-08-27 00:54:46 +08:00
|
|
|
DO idx = 1, ntgrp
|
2016-02-14 20:49:16 +08:00
|
|
|
!
|
2016-07-29 22:25:22 +08:00
|
|
|
! ... dtgs%nogrp ffts at the same time. We prepare both
|
2016-02-14 20:49:16 +08:00
|
|
|
! evc (at k) and dpsi (at k+q)
|
|
|
|
!
|
|
|
|
IF( idx + ibnd - 1 <= nbnd_occ(ikk) ) THEN
|
|
|
|
!
|
|
|
|
DO ig = 1, npw
|
2017-12-24 22:24:26 +08:00
|
|
|
tg_psi( dffts%nl( igk_k( ig,ikk ) ) + ioff, 1 ) = evc( ig, idx+ibnd-1 )
|
|
|
|
tg_psi( dffts%nl( igk_k( ig,ikk ) ) + ioff, 2 ) = evc( npwx+ig, idx+ibnd-1 )
|
2016-02-14 20:49:16 +08:00
|
|
|
END DO
|
|
|
|
DO ig = 1, npwq
|
2017-12-24 22:24:26 +08:00
|
|
|
tg_dpsi( dffts%nl( igk_k( ig,ikq ) ) + ioff, 1 ) = dpsi( ig, idx+ibnd-1 )
|
|
|
|
tg_dpsi( dffts%nl( igk_k( ig,ikq ) ) + ioff, 2 ) = dpsi( npwx+ig, idx+ibnd-1 )
|
2016-02-14 20:49:16 +08:00
|
|
|
END DO
|
|
|
|
!
|
|
|
|
END IF
|
|
|
|
!
|
2017-08-27 00:54:46 +08:00
|
|
|
ioff = ioff + right_inc
|
2016-02-14 20:49:16 +08:00
|
|
|
!
|
|
|
|
END DO
|
MAJOR restructuring of the FFTXlib library
In real space processors are organized in a 2D pattern.
Each processor owns data from a sub-set of Z-planes and a sub-set of Y-planes.
In reciprocal space each processor owns Z-columns that belong to a sub set of
X-values. This allows to split the processors in two sets for communication
in the YZ and XY planes.
In alternative, if the situation allows for it, a task group paralelization is used
(with ntg=nyfft) where complete XY planes of ntg wavefunctions are collected and Fourier
trasnformed in G space by different task-groups. This is preferable to the Z-proc + Y-proc
paralleization if task group can be used because a smaller number of larger ammounts of
data are transferred. Hence three types of fft are implemented:
!
!! ... isgn = +-1 : parallel 3d fft for rho and for the potential
!
!! ... isgn = +-2 : parallel 3d fft for wavefunctions
!
!! ... isgn = +-3 : parallel 3d fft for wavefunctions with task group
!
!! ... isgn = + : G-space to R-space, output = \sum_G f(G)exp(+iG*R)
!! ... fft along z using pencils (cft_1z)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along x using pencils (cft_1x)
!
!! ... isgn = - : R-space to G-space, output = \int_R f(R)exp(-iG*R)/Omega
!! ... fft along x using pencils (cft_1x)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along z using pencils (cft_1z)
!
! If task_group_fft_is_active the FFT acts on a number of wfcs equal to
! dfft%nproc2, the number of Y-sections in which a plane is divided.
! Data are reshuffled by the fft_scatter_tg routine so that each of the
! dfft%nproc2 subgroups (made by dfft%nproc3 procs) deals with whole planes
! of a single wavefunciton.
!
fft_type module heavily modified, a number of variables renamed with more intuitive names
(at least to me), a number of more variables introduced for the Y-proc parallelization.
Task_group module made void. task_group management is now reduced to the logical component
fft_desc%have_task_groups of fft_type_descriptor type variable fft_desc.
In term of interfaces, the 'easy' calling sequences are
SUBROUTINE invfft/fwfft( grid_type, f, dfft, howmany )
!! where:
!!
!! **grid_type = 'Dense'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the dense grid (dfftp). On output, f is overwritten
!!
!! **grid_type = 'Smooth'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Wave'** :
!! inverse/direct fourier transform of wave functions f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'tgWave'** :
!! inverse/direct fourier transform of wave functions f with task group
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Custom'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **grid_type = 'CustomWave'** :
!! inverse/direct fourier transform of wave functions f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **dfft = FFT descriptor**, IMPORTANT NOTICE: grid is specified only by dfft.
!! No check is performed on the correspondence between dfft and grid_type.
!! grid_type is now used only to distinguish cases 'Wave' / 'CustomWave'
!! from all other cases
Many more files modified.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@13676 c92efa57-630b-4861-b058-cf58834340f0
2017-08-02 04:31:02 +08:00
|
|
|
CALL invfft ('tgWave', tg_psi(:,1), dffts)
|
|
|
|
CALL invfft ('tgWave', tg_psi(:,2), dffts)
|
|
|
|
CALL invfft ('tgWave', tg_dpsi(:,1), dffts)
|
|
|
|
CALL invfft ('tgWave', tg_dpsi(:,2), dffts)
|
2016-02-14 20:49:16 +08:00
|
|
|
|
MAJOR restructuring of the FFTXlib library
In real space processors are organized in a 2D pattern.
Each processor owns data from a sub-set of Z-planes and a sub-set of Y-planes.
In reciprocal space each processor owns Z-columns that belong to a sub set of
X-values. This allows to split the processors in two sets for communication
in the YZ and XY planes.
In alternative, if the situation allows for it, a task group paralelization is used
(with ntg=nyfft) where complete XY planes of ntg wavefunctions are collected and Fourier
trasnformed in G space by different task-groups. This is preferable to the Z-proc + Y-proc
paralleization if task group can be used because a smaller number of larger ammounts of
data are transferred. Hence three types of fft are implemented:
!
!! ... isgn = +-1 : parallel 3d fft for rho and for the potential
!
!! ... isgn = +-2 : parallel 3d fft for wavefunctions
!
!! ... isgn = +-3 : parallel 3d fft for wavefunctions with task group
!
!! ... isgn = + : G-space to R-space, output = \sum_G f(G)exp(+iG*R)
!! ... fft along z using pencils (cft_1z)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along x using pencils (cft_1x)
!
!! ... isgn = - : R-space to G-space, output = \int_R f(R)exp(-iG*R)/Omega
!! ... fft along x using pencils (cft_1x)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along z using pencils (cft_1z)
!
! If task_group_fft_is_active the FFT acts on a number of wfcs equal to
! dfft%nproc2, the number of Y-sections in which a plane is divided.
! Data are reshuffled by the fft_scatter_tg routine so that each of the
! dfft%nproc2 subgroups (made by dfft%nproc3 procs) deals with whole planes
! of a single wavefunciton.
!
fft_type module heavily modified, a number of variables renamed with more intuitive names
(at least to me), a number of more variables introduced for the Y-proc parallelization.
Task_group module made void. task_group management is now reduced to the logical component
fft_desc%have_task_groups of fft_type_descriptor type variable fft_desc.
In term of interfaces, the 'easy' calling sequences are
SUBROUTINE invfft/fwfft( grid_type, f, dfft, howmany )
!! where:
!!
!! **grid_type = 'Dense'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the dense grid (dfftp). On output, f is overwritten
!!
!! **grid_type = 'Smooth'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Wave'** :
!! inverse/direct fourier transform of wave functions f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'tgWave'** :
!! inverse/direct fourier transform of wave functions f with task group
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Custom'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **grid_type = 'CustomWave'** :
!! inverse/direct fourier transform of wave functions f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **dfft = FFT descriptor**, IMPORTANT NOTICE: grid is specified only by dfft.
!! No check is performed on the correspondence between dfft and grid_type.
!! grid_type is now used only to distinguish cases 'Wave' / 'CustomWave'
!! from all other cases
Many more files modified.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@13676 c92efa57-630b-4861-b058-cf58834340f0
2017-08-02 04:31:02 +08:00
|
|
|
do ir = 1, dffts%nr1x * dffts%nr2x * dffts%my_nr3p
|
|
|
|
tg_drho (ir,1) = tg_drho (ir,1) + wgt * (CONJG(tg_psi (ir,1) )*tg_dpsi (ir,1) &
|
|
|
|
+ CONJG(tg_psi (ir,2) )*tg_dpsi (ir,2) )
|
2016-02-14 20:49:16 +08:00
|
|
|
enddo
|
|
|
|
IF (domag) THEN
|
MAJOR restructuring of the FFTXlib library
In real space processors are organized in a 2D pattern.
Each processor owns data from a sub-set of Z-planes and a sub-set of Y-planes.
In reciprocal space each processor owns Z-columns that belong to a sub set of
X-values. This allows to split the processors in two sets for communication
in the YZ and XY planes.
In alternative, if the situation allows for it, a task group paralelization is used
(with ntg=nyfft) where complete XY planes of ntg wavefunctions are collected and Fourier
trasnformed in G space by different task-groups. This is preferable to the Z-proc + Y-proc
paralleization if task group can be used because a smaller number of larger ammounts of
data are transferred. Hence three types of fft are implemented:
!
!! ... isgn = +-1 : parallel 3d fft for rho and for the potential
!
!! ... isgn = +-2 : parallel 3d fft for wavefunctions
!
!! ... isgn = +-3 : parallel 3d fft for wavefunctions with task group
!
!! ... isgn = + : G-space to R-space, output = \sum_G f(G)exp(+iG*R)
!! ... fft along z using pencils (cft_1z)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along x using pencils (cft_1x)
!
!! ... isgn = - : R-space to G-space, output = \int_R f(R)exp(-iG*R)/Omega
!! ... fft along x using pencils (cft_1x)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along z using pencils (cft_1z)
!
! If task_group_fft_is_active the FFT acts on a number of wfcs equal to
! dfft%nproc2, the number of Y-sections in which a plane is divided.
! Data are reshuffled by the fft_scatter_tg routine so that each of the
! dfft%nproc2 subgroups (made by dfft%nproc3 procs) deals with whole planes
! of a single wavefunciton.
!
fft_type module heavily modified, a number of variables renamed with more intuitive names
(at least to me), a number of more variables introduced for the Y-proc parallelization.
Task_group module made void. task_group management is now reduced to the logical component
fft_desc%have_task_groups of fft_type_descriptor type variable fft_desc.
In term of interfaces, the 'easy' calling sequences are
SUBROUTINE invfft/fwfft( grid_type, f, dfft, howmany )
!! where:
!!
!! **grid_type = 'Dense'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the dense grid (dfftp). On output, f is overwritten
!!
!! **grid_type = 'Smooth'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Wave'** :
!! inverse/direct fourier transform of wave functions f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'tgWave'** :
!! inverse/direct fourier transform of wave functions f with task group
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Custom'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **grid_type = 'CustomWave'** :
!! inverse/direct fourier transform of wave functions f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **dfft = FFT descriptor**, IMPORTANT NOTICE: grid is specified only by dfft.
!! No check is performed on the correspondence between dfft and grid_type.
!! grid_type is now used only to distinguish cases 'Wave' / 'CustomWave'
!! from all other cases
Many more files modified.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@13676 c92efa57-630b-4861-b058-cf58834340f0
2017-08-02 04:31:02 +08:00
|
|
|
do ir = 1, dffts%nr1x * dffts%nr2x * dffts%my_nr3p
|
2020-07-04 01:40:46 +08:00
|
|
|
tg_drho(ir,2)= tg_drho(ir,2) + (rsign) *wgt * (CONJG(tg_psi(ir,1))*tg_dpsi(ir,2) &
|
MAJOR restructuring of the FFTXlib library
In real space processors are organized in a 2D pattern.
Each processor owns data from a sub-set of Z-planes and a sub-set of Y-planes.
In reciprocal space each processor owns Z-columns that belong to a sub set of
X-values. This allows to split the processors in two sets for communication
in the YZ and XY planes.
In alternative, if the situation allows for it, a task group paralelization is used
(with ntg=nyfft) where complete XY planes of ntg wavefunctions are collected and Fourier
trasnformed in G space by different task-groups. This is preferable to the Z-proc + Y-proc
paralleization if task group can be used because a smaller number of larger ammounts of
data are transferred. Hence three types of fft are implemented:
!
!! ... isgn = +-1 : parallel 3d fft for rho and for the potential
!
!! ... isgn = +-2 : parallel 3d fft for wavefunctions
!
!! ... isgn = +-3 : parallel 3d fft for wavefunctions with task group
!
!! ... isgn = + : G-space to R-space, output = \sum_G f(G)exp(+iG*R)
!! ... fft along z using pencils (cft_1z)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along x using pencils (cft_1x)
!
!! ... isgn = - : R-space to G-space, output = \int_R f(R)exp(-iG*R)/Omega
!! ... fft along x using pencils (cft_1x)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along z using pencils (cft_1z)
!
! If task_group_fft_is_active the FFT acts on a number of wfcs equal to
! dfft%nproc2, the number of Y-sections in which a plane is divided.
! Data are reshuffled by the fft_scatter_tg routine so that each of the
! dfft%nproc2 subgroups (made by dfft%nproc3 procs) deals with whole planes
! of a single wavefunciton.
!
fft_type module heavily modified, a number of variables renamed with more intuitive names
(at least to me), a number of more variables introduced for the Y-proc parallelization.
Task_group module made void. task_group management is now reduced to the logical component
fft_desc%have_task_groups of fft_type_descriptor type variable fft_desc.
In term of interfaces, the 'easy' calling sequences are
SUBROUTINE invfft/fwfft( grid_type, f, dfft, howmany )
!! where:
!!
!! **grid_type = 'Dense'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the dense grid (dfftp). On output, f is overwritten
!!
!! **grid_type = 'Smooth'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Wave'** :
!! inverse/direct fourier transform of wave functions f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'tgWave'** :
!! inverse/direct fourier transform of wave functions f with task group
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Custom'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **grid_type = 'CustomWave'** :
!! inverse/direct fourier transform of wave functions f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **dfft = FFT descriptor**, IMPORTANT NOTICE: grid is specified only by dfft.
!! No check is performed on the correspondence between dfft and grid_type.
!! grid_type is now used only to distinguish cases 'Wave' / 'CustomWave'
!! from all other cases
Many more files modified.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@13676 c92efa57-630b-4861-b058-cf58834340f0
2017-08-02 04:31:02 +08:00
|
|
|
+ CONJG(tg_psi(ir,2))*tg_dpsi(ir,1) )
|
2020-07-04 01:40:46 +08:00
|
|
|
tg_drho(ir,3)= tg_drho(ir,3) + (rsign) *wgt * (CONJG(tg_psi(ir,1))*tg_dpsi(ir,2) &
|
MAJOR restructuring of the FFTXlib library
In real space processors are organized in a 2D pattern.
Each processor owns data from a sub-set of Z-planes and a sub-set of Y-planes.
In reciprocal space each processor owns Z-columns that belong to a sub set of
X-values. This allows to split the processors in two sets for communication
in the YZ and XY planes.
In alternative, if the situation allows for it, a task group paralelization is used
(with ntg=nyfft) where complete XY planes of ntg wavefunctions are collected and Fourier
trasnformed in G space by different task-groups. This is preferable to the Z-proc + Y-proc
paralleization if task group can be used because a smaller number of larger ammounts of
data are transferred. Hence three types of fft are implemented:
!
!! ... isgn = +-1 : parallel 3d fft for rho and for the potential
!
!! ... isgn = +-2 : parallel 3d fft for wavefunctions
!
!! ... isgn = +-3 : parallel 3d fft for wavefunctions with task group
!
!! ... isgn = + : G-space to R-space, output = \sum_G f(G)exp(+iG*R)
!! ... fft along z using pencils (cft_1z)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along x using pencils (cft_1x)
!
!! ... isgn = - : R-space to G-space, output = \int_R f(R)exp(-iG*R)/Omega
!! ... fft along x using pencils (cft_1x)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along z using pencils (cft_1z)
!
! If task_group_fft_is_active the FFT acts on a number of wfcs equal to
! dfft%nproc2, the number of Y-sections in which a plane is divided.
! Data are reshuffled by the fft_scatter_tg routine so that each of the
! dfft%nproc2 subgroups (made by dfft%nproc3 procs) deals with whole planes
! of a single wavefunciton.
!
fft_type module heavily modified, a number of variables renamed with more intuitive names
(at least to me), a number of more variables introduced for the Y-proc parallelization.
Task_group module made void. task_group management is now reduced to the logical component
fft_desc%have_task_groups of fft_type_descriptor type variable fft_desc.
In term of interfaces, the 'easy' calling sequences are
SUBROUTINE invfft/fwfft( grid_type, f, dfft, howmany )
!! where:
!!
!! **grid_type = 'Dense'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the dense grid (dfftp). On output, f is overwritten
!!
!! **grid_type = 'Smooth'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Wave'** :
!! inverse/direct fourier transform of wave functions f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'tgWave'** :
!! inverse/direct fourier transform of wave functions f with task group
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Custom'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **grid_type = 'CustomWave'** :
!! inverse/direct fourier transform of wave functions f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **dfft = FFT descriptor**, IMPORTANT NOTICE: grid is specified only by dfft.
!! No check is performed on the correspondence between dfft and grid_type.
!! grid_type is now used only to distinguish cases 'Wave' / 'CustomWave'
!! from all other cases
Many more files modified.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@13676 c92efa57-630b-4861-b058-cf58834340f0
2017-08-02 04:31:02 +08:00
|
|
|
- CONJG(tg_psi(ir,2))*tg_dpsi(ir,1) ) * (0.d0,-1.d0)
|
2020-07-04 01:40:46 +08:00
|
|
|
tg_drho(ir,4)= tg_drho(ir,4) + (rsign) *wgt * (CONJG(tg_psi(ir,1))*tg_dpsi(ir,1) &
|
MAJOR restructuring of the FFTXlib library
In real space processors are organized in a 2D pattern.
Each processor owns data from a sub-set of Z-planes and a sub-set of Y-planes.
In reciprocal space each processor owns Z-columns that belong to a sub set of
X-values. This allows to split the processors in two sets for communication
in the YZ and XY planes.
In alternative, if the situation allows for it, a task group paralelization is used
(with ntg=nyfft) where complete XY planes of ntg wavefunctions are collected and Fourier
trasnformed in G space by different task-groups. This is preferable to the Z-proc + Y-proc
paralleization if task group can be used because a smaller number of larger ammounts of
data are transferred. Hence three types of fft are implemented:
!
!! ... isgn = +-1 : parallel 3d fft for rho and for the potential
!
!! ... isgn = +-2 : parallel 3d fft for wavefunctions
!
!! ... isgn = +-3 : parallel 3d fft for wavefunctions with task group
!
!! ... isgn = + : G-space to R-space, output = \sum_G f(G)exp(+iG*R)
!! ... fft along z using pencils (cft_1z)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along x using pencils (cft_1x)
!
!! ... isgn = - : R-space to G-space, output = \int_R f(R)exp(-iG*R)/Omega
!! ... fft along x using pencils (cft_1x)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along z using pencils (cft_1z)
!
! If task_group_fft_is_active the FFT acts on a number of wfcs equal to
! dfft%nproc2, the number of Y-sections in which a plane is divided.
! Data are reshuffled by the fft_scatter_tg routine so that each of the
! dfft%nproc2 subgroups (made by dfft%nproc3 procs) deals with whole planes
! of a single wavefunciton.
!
fft_type module heavily modified, a number of variables renamed with more intuitive names
(at least to me), a number of more variables introduced for the Y-proc parallelization.
Task_group module made void. task_group management is now reduced to the logical component
fft_desc%have_task_groups of fft_type_descriptor type variable fft_desc.
In term of interfaces, the 'easy' calling sequences are
SUBROUTINE invfft/fwfft( grid_type, f, dfft, howmany )
!! where:
!!
!! **grid_type = 'Dense'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the dense grid (dfftp). On output, f is overwritten
!!
!! **grid_type = 'Smooth'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Wave'** :
!! inverse/direct fourier transform of wave functions f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'tgWave'** :
!! inverse/direct fourier transform of wave functions f with task group
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Custom'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **grid_type = 'CustomWave'** :
!! inverse/direct fourier transform of wave functions f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **dfft = FFT descriptor**, IMPORTANT NOTICE: grid is specified only by dfft.
!! No check is performed on the correspondence between dfft and grid_type.
!! grid_type is now used only to distinguish cases 'Wave' / 'CustomWave'
!! from all other cases
Many more files modified.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@13676 c92efa57-630b-4861-b058-cf58834340f0
2017-08-02 04:31:02 +08:00
|
|
|
- CONJG(tg_psi(ir,2))*tg_dpsi(ir,2) )
|
2016-02-14 20:49:16 +08:00
|
|
|
enddo
|
|
|
|
ENDIF
|
|
|
|
!
|
|
|
|
! reduce the group charge (equivalent to sum over the bands of the
|
|
|
|
! orbital group)
|
|
|
|
!
|
2017-08-27 00:54:46 +08:00
|
|
|
CALL tg_reduce_rho( drhoscf, tg_drho, dffts )
|
2016-02-14 20:49:16 +08:00
|
|
|
!
|
|
|
|
ELSE
|
|
|
|
!
|
|
|
|
! Normal case: no task groups
|
|
|
|
!
|
|
|
|
! FFT to R-space of the unperturbed wfct's evc
|
|
|
|
!
|
|
|
|
psi = (0.d0, 0.d0)
|
|
|
|
do ig = 1, npw
|
2017-12-24 22:24:26 +08:00
|
|
|
psi (dffts%nl (igk_k(ig,ikk) ), 1) = evc (ig, ibnd)
|
|
|
|
psi (dffts%nl (igk_k(ig,ikk) ), 2) = evc (ig+npwx, ibnd)
|
2016-02-14 20:49:16 +08:00
|
|
|
enddo
|
|
|
|
CALL invfft ('Wave', psi(:,1), dffts)
|
|
|
|
CALL invfft ('Wave', psi(:,2), dffts)
|
|
|
|
!
|
|
|
|
! FFT to R-space of the perturbed wfct's dpsi
|
|
|
|
!
|
|
|
|
dpsic = (0.d0, 0.d0)
|
|
|
|
do ig = 1, npwq
|
2017-12-24 22:24:26 +08:00
|
|
|
dpsic (dffts%nl (igk_k(ig,ikq)), 1 ) = dpsi (ig, ibnd)
|
|
|
|
dpsic (dffts%nl (igk_k(ig,ikq)), 2 ) = dpsi (ig+npwx, ibnd)
|
2016-02-14 20:49:16 +08:00
|
|
|
enddo
|
|
|
|
CALL invfft ('Wave', dpsic(:,1), dffts)
|
|
|
|
CALL invfft ('Wave', dpsic(:,2), dffts)
|
|
|
|
!
|
|
|
|
! Calculation of the response charge density
|
|
|
|
!
|
|
|
|
do ir = 1, dffts%nnr
|
|
|
|
drhoscf(ir,1)=drhoscf(ir,1)+wgt*(CONJG(psi(ir,1))*dpsic(ir,1) + &
|
|
|
|
CONJG(psi(ir,2))*dpsic(ir,2) )
|
|
|
|
enddo
|
|
|
|
IF (domag) THEN
|
|
|
|
do ir = 1, dffts%nnr
|
2020-07-04 01:40:46 +08:00
|
|
|
drhoscf(ir,2)=drhoscf (ir,2) + (rsign) *wgt * (CONJG(psi(ir,1))*dpsic(ir,2) &
|
MAJOR restructuring of the FFTXlib library
In real space processors are organized in a 2D pattern.
Each processor owns data from a sub-set of Z-planes and a sub-set of Y-planes.
In reciprocal space each processor owns Z-columns that belong to a sub set of
X-values. This allows to split the processors in two sets for communication
in the YZ and XY planes.
In alternative, if the situation allows for it, a task group paralelization is used
(with ntg=nyfft) where complete XY planes of ntg wavefunctions are collected and Fourier
trasnformed in G space by different task-groups. This is preferable to the Z-proc + Y-proc
paralleization if task group can be used because a smaller number of larger ammounts of
data are transferred. Hence three types of fft are implemented:
!
!! ... isgn = +-1 : parallel 3d fft for rho and for the potential
!
!! ... isgn = +-2 : parallel 3d fft for wavefunctions
!
!! ... isgn = +-3 : parallel 3d fft for wavefunctions with task group
!
!! ... isgn = + : G-space to R-space, output = \sum_G f(G)exp(+iG*R)
!! ... fft along z using pencils (cft_1z)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along x using pencils (cft_1x)
!
!! ... isgn = - : R-space to G-space, output = \int_R f(R)exp(-iG*R)/Omega
!! ... fft along x using pencils (cft_1x)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along z using pencils (cft_1z)
!
! If task_group_fft_is_active the FFT acts on a number of wfcs equal to
! dfft%nproc2, the number of Y-sections in which a plane is divided.
! Data are reshuffled by the fft_scatter_tg routine so that each of the
! dfft%nproc2 subgroups (made by dfft%nproc3 procs) deals with whole planes
! of a single wavefunciton.
!
fft_type module heavily modified, a number of variables renamed with more intuitive names
(at least to me), a number of more variables introduced for the Y-proc parallelization.
Task_group module made void. task_group management is now reduced to the logical component
fft_desc%have_task_groups of fft_type_descriptor type variable fft_desc.
In term of interfaces, the 'easy' calling sequences are
SUBROUTINE invfft/fwfft( grid_type, f, dfft, howmany )
!! where:
!!
!! **grid_type = 'Dense'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the dense grid (dfftp). On output, f is overwritten
!!
!! **grid_type = 'Smooth'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Wave'** :
!! inverse/direct fourier transform of wave functions f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'tgWave'** :
!! inverse/direct fourier transform of wave functions f with task group
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Custom'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **grid_type = 'CustomWave'** :
!! inverse/direct fourier transform of wave functions f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **dfft = FFT descriptor**, IMPORTANT NOTICE: grid is specified only by dfft.
!! No check is performed on the correspondence between dfft and grid_type.
!! grid_type is now used only to distinguish cases 'Wave' / 'CustomWave'
!! from all other cases
Many more files modified.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@13676 c92efa57-630b-4861-b058-cf58834340f0
2017-08-02 04:31:02 +08:00
|
|
|
+ CONJG(psi(ir,2))*dpsic(ir,1) )
|
2020-07-04 01:40:46 +08:00
|
|
|
drhoscf(ir,3)=drhoscf (ir,3) + (rsign) *wgt * (CONJG(psi(ir,1))*dpsic(ir,2) &
|
MAJOR restructuring of the FFTXlib library
In real space processors are organized in a 2D pattern.
Each processor owns data from a sub-set of Z-planes and a sub-set of Y-planes.
In reciprocal space each processor owns Z-columns that belong to a sub set of
X-values. This allows to split the processors in two sets for communication
in the YZ and XY planes.
In alternative, if the situation allows for it, a task group paralelization is used
(with ntg=nyfft) where complete XY planes of ntg wavefunctions are collected and Fourier
trasnformed in G space by different task-groups. This is preferable to the Z-proc + Y-proc
paralleization if task group can be used because a smaller number of larger ammounts of
data are transferred. Hence three types of fft are implemented:
!
!! ... isgn = +-1 : parallel 3d fft for rho and for the potential
!
!! ... isgn = +-2 : parallel 3d fft for wavefunctions
!
!! ... isgn = +-3 : parallel 3d fft for wavefunctions with task group
!
!! ... isgn = + : G-space to R-space, output = \sum_G f(G)exp(+iG*R)
!! ... fft along z using pencils (cft_1z)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along x using pencils (cft_1x)
!
!! ... isgn = - : R-space to G-space, output = \int_R f(R)exp(-iG*R)/Omega
!! ... fft along x using pencils (cft_1x)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along z using pencils (cft_1z)
!
! If task_group_fft_is_active the FFT acts on a number of wfcs equal to
! dfft%nproc2, the number of Y-sections in which a plane is divided.
! Data are reshuffled by the fft_scatter_tg routine so that each of the
! dfft%nproc2 subgroups (made by dfft%nproc3 procs) deals with whole planes
! of a single wavefunciton.
!
fft_type module heavily modified, a number of variables renamed with more intuitive names
(at least to me), a number of more variables introduced for the Y-proc parallelization.
Task_group module made void. task_group management is now reduced to the logical component
fft_desc%have_task_groups of fft_type_descriptor type variable fft_desc.
In term of interfaces, the 'easy' calling sequences are
SUBROUTINE invfft/fwfft( grid_type, f, dfft, howmany )
!! where:
!!
!! **grid_type = 'Dense'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the dense grid (dfftp). On output, f is overwritten
!!
!! **grid_type = 'Smooth'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Wave'** :
!! inverse/direct fourier transform of wave functions f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'tgWave'** :
!! inverse/direct fourier transform of wave functions f with task group
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Custom'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **grid_type = 'CustomWave'** :
!! inverse/direct fourier transform of wave functions f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **dfft = FFT descriptor**, IMPORTANT NOTICE: grid is specified only by dfft.
!! No check is performed on the correspondence between dfft and grid_type.
!! grid_type is now used only to distinguish cases 'Wave' / 'CustomWave'
!! from all other cases
Many more files modified.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@13676 c92efa57-630b-4861-b058-cf58834340f0
2017-08-02 04:31:02 +08:00
|
|
|
- CONJG(psi(ir,2))*dpsic(ir,1) ) * (0.d0,-1.d0)
|
2020-07-04 01:40:46 +08:00
|
|
|
drhoscf(ir,4)=drhoscf (ir,4) + (rsign) *wgt * (CONJG(psi(ir,1))*dpsic(ir,1) &
|
MAJOR restructuring of the FFTXlib library
In real space processors are organized in a 2D pattern.
Each processor owns data from a sub-set of Z-planes and a sub-set of Y-planes.
In reciprocal space each processor owns Z-columns that belong to a sub set of
X-values. This allows to split the processors in two sets for communication
in the YZ and XY planes.
In alternative, if the situation allows for it, a task group paralelization is used
(with ntg=nyfft) where complete XY planes of ntg wavefunctions are collected and Fourier
trasnformed in G space by different task-groups. This is preferable to the Z-proc + Y-proc
paralleization if task group can be used because a smaller number of larger ammounts of
data are transferred. Hence three types of fft are implemented:
!
!! ... isgn = +-1 : parallel 3d fft for rho and for the potential
!
!! ... isgn = +-2 : parallel 3d fft for wavefunctions
!
!! ... isgn = +-3 : parallel 3d fft for wavefunctions with task group
!
!! ... isgn = + : G-space to R-space, output = \sum_G f(G)exp(+iG*R)
!! ... fft along z using pencils (cft_1z)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along x using pencils (cft_1x)
!
!! ... isgn = - : R-space to G-space, output = \int_R f(R)exp(-iG*R)/Omega
!! ... fft along x using pencils (cft_1x)
!! ... transpose across nodes (fft_scatter_xy)
!! ... fft along y using pencils (cft_1y)
!! ... transpose across nodes (fft_scatter_yz)
!! ... fft along z using pencils (cft_1z)
!
! If task_group_fft_is_active the FFT acts on a number of wfcs equal to
! dfft%nproc2, the number of Y-sections in which a plane is divided.
! Data are reshuffled by the fft_scatter_tg routine so that each of the
! dfft%nproc2 subgroups (made by dfft%nproc3 procs) deals with whole planes
! of a single wavefunciton.
!
fft_type module heavily modified, a number of variables renamed with more intuitive names
(at least to me), a number of more variables introduced for the Y-proc parallelization.
Task_group module made void. task_group management is now reduced to the logical component
fft_desc%have_task_groups of fft_type_descriptor type variable fft_desc.
In term of interfaces, the 'easy' calling sequences are
SUBROUTINE invfft/fwfft( grid_type, f, dfft, howmany )
!! where:
!!
!! **grid_type = 'Dense'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the dense grid (dfftp). On output, f is overwritten
!!
!! **grid_type = 'Smooth'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Wave'** :
!! inverse/direct fourier transform of wave functions f
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'tgWave'** :
!! inverse/direct fourier transform of wave functions f with task group
!! on the smooth grid (dffts). On output, f is overwritten
!!
!! **grid_type = 'Custom'** :
!! inverse/direct fourier transform of potentials and charge density f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **grid_type = 'CustomWave'** :
!! inverse/direct fourier transform of wave functions f
!! on a custom grid (dfft_exx). On output, f is overwritten
!!
!! **dfft = FFT descriptor**, IMPORTANT NOTICE: grid is specified only by dfft.
!! No check is performed on the correspondence between dfft and grid_type.
!! grid_type is now used only to distinguish cases 'Wave' / 'CustomWave'
!! from all other cases
Many more files modified.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@13676 c92efa57-630b-4861-b058-cf58834340f0
2017-08-02 04:31:02 +08:00
|
|
|
- CONJG(psi(ir,2))*dpsic(ir,2) )
|
2016-02-14 20:49:16 +08:00
|
|
|
enddo
|
|
|
|
END IF
|
|
|
|
!
|
|
|
|
END IF
|
|
|
|
!
|
|
|
|
enddo
|
|
|
|
!
|
|
|
|
! Ultrasoft contribution
|
|
|
|
! Calculate dbecsum_nc = <evc|vkb><vkb|dpsi>
|
|
|
|
!
|
2020-07-04 01:40:46 +08:00
|
|
|
IF (rsign==1.0d0) THEN
|
2020-01-29 18:44:02 +08:00
|
|
|
CALL addusdbec_nc (ik, weight, dpsi, dbecsum, becp1)
|
|
|
|
ELSE
|
|
|
|
CALL addusdbec_nc (ik, weight, dpsi, dbecsum, becpt)
|
|
|
|
ENDIF
|
2016-02-14 20:49:16 +08:00
|
|
|
!
|
|
|
|
DEALLOCATE(psi)
|
|
|
|
DEALLOCATE(dpsic)
|
|
|
|
!
|
2018-01-03 08:38:48 +08:00
|
|
|
IF (dffts%has_task_groups) THEN
|
2016-02-14 20:49:16 +08:00
|
|
|
DEALLOCATE(tg_psi)
|
|
|
|
DEALLOCATE(tg_dpsi)
|
|
|
|
DEALLOCATE(tg_drho)
|
|
|
|
END IF
|
|
|
|
!
|
|
|
|
CALL stop_clock ('incdrhoscf')
|
|
|
|
!
|
|
|
|
RETURN
|
|
|
|
!
|
|
|
|
end subroutine incdrhoscf_nc
|