mirror of
https://github.com/TREX-CoE/irpjast.git
synced 2024-11-03 20:54:10 +01:00
Optimizations: reduce number of exponentials
This commit is contained in:
parent
c2bbae93e7
commit
7fd1fd19fd
7
Makefile
7
Makefile
@ -1,8 +1,9 @@
|
|||||||
IRPF90 = irpf90 #-a -d
|
IRPF90 = irpf90 --codelet=factor_een:100000 #-a -d
|
||||||
FC = gfortran
|
FC = ifort -xHost -g
|
||||||
FCFLAGS= -O2 -ffree-line-length-none -I .
|
FCFLAGS= -O2 -ffree-line-length-none -I .
|
||||||
NINJA = ninja
|
NINJA = ninja
|
||||||
AR = ar
|
AR = ar
|
||||||
|
ARCHIVE = ar crs
|
||||||
RANLIB = ranlib
|
RANLIB = ranlib
|
||||||
|
|
||||||
SRC=
|
SRC=
|
||||||
|
@ -4,7 +4,7 @@ BEGIN_PROVIDER [ double precision, factor_een ]
|
|||||||
! ElectronE-electron-nuclei contribution to Jastrow factor
|
! ElectronE-electron-nuclei contribution to Jastrow factor
|
||||||
END_DOC
|
END_DOC
|
||||||
integer :: i, j, a, p, k, l, lmax, m
|
integer :: i, j, a, p, k, l, lmax, m
|
||||||
double precision :: riam, rjam_cn, rial, rjal, rijk
|
double precision :: rjam_cn
|
||||||
double precision :: cn
|
double precision :: cn
|
||||||
|
|
||||||
factor_een = 0.0d0
|
factor_een = 0.0d0
|
||||||
@ -21,14 +21,16 @@ BEGIN_PROVIDER [ double precision, factor_een ]
|
|||||||
m = (p - k - l) / 2
|
m = (p - k - l) / 2
|
||||||
do a = 1, nnuc
|
do a = 1, nnuc
|
||||||
cn = cord_vect_lkp(l, k, p, typenuc_arr(a))
|
cn = cord_vect_lkp(l, k, p, typenuc_arr(a))
|
||||||
do j = 1, nelec
|
rjam_cn = rescale_een_n(2, a, m) * cn
|
||||||
rjal = rescale_een_n(j, a, l)
|
factor_een = factor_een + rescale_een_e(1,2,k) * &
|
||||||
|
(rescale_een_n(1,a,l) + rescale_een_n(2,a,l)) * &
|
||||||
|
rescale_een_n(1,a,m) * rjam_cn
|
||||||
|
do j = 3, nelec
|
||||||
rjam_cn = rescale_een_n(j, a, m) * cn
|
rjam_cn = rescale_een_n(j, a, m) * cn
|
||||||
do i = 1, j - 1
|
do i = 1, j - 1
|
||||||
rial = rescale_een_n(i, a, l)
|
factor_een = factor_een + rescale_een_e(i,j,k) * &
|
||||||
riam = rescale_een_n(i, a, m)
|
(rescale_een_n(i,a,l) + rescale_een_n(j,a,l)) * &
|
||||||
rijk = rescale_een_e(i, j, k)
|
rescale_een_n(i,a,m) * rjam_cn
|
||||||
factor_een = factor_een + rijk * (rial + rjal) * riam * rjam_cn
|
|
||||||
enddo
|
enddo
|
||||||
enddo
|
enddo
|
||||||
enddo
|
enddo
|
||||||
|
@ -20,11 +20,15 @@ BEGIN_PROVIDER [ double precision, rescale_ee, (nelec, nelec) ]
|
|||||||
! R = (1 - exp(-kappa r))/kappa for electron-electron for $J_{ee}$
|
! R = (1 - exp(-kappa r))/kappa for electron-electron for $J_{ee}$
|
||||||
END_DOC
|
END_DOC
|
||||||
integer :: i, j
|
integer :: i, j
|
||||||
|
double precision :: x
|
||||||
|
|
||||||
do j = 1, nelec
|
do j = 1, nelec
|
||||||
do i = 1, nelec
|
do i = 1, j-1
|
||||||
rescale_ee(i, j) = (1.0d0 - dexp(-kappa * elec_dist(i, j))) * kappa_inv
|
x = (1.0d0 - dexp(-kappa * elec_dist(i, j))) * kappa_inv
|
||||||
|
rescale_ee(i, j) = x
|
||||||
|
rescale_ee(j, i) = x
|
||||||
enddo
|
enddo
|
||||||
|
rescale_ee(j, j) = 0.d0
|
||||||
enddo
|
enddo
|
||||||
END_PROVIDER
|
END_PROVIDER
|
||||||
|
|
||||||
@ -36,9 +40,11 @@ BEGIN_PROVIDER [ double precision, rescale_ee_deriv_e, (4, nelec, nelec) ]
|
|||||||
! Dimension 4 : d2x + d2y + d2z
|
! Dimension 4 : d2x + d2y + d2z
|
||||||
END_DOC
|
END_DOC
|
||||||
integer :: i, j, ii
|
integer :: i, j, ii
|
||||||
|
double precision :: f
|
||||||
|
|
||||||
do j = 1, nelec
|
do j = 1, nelec
|
||||||
do i = 1, nelec
|
do i = 1, nelec
|
||||||
|
f = 1.d0 - kappa*rescale_ee(i,j) ! == dexp(-kappa * elec_dist(i, j))
|
||||||
do ii = 1, 4
|
do ii = 1, 4
|
||||||
rescale_ee_deriv_e(ii, i, j) = elec_dist_deriv_e(ii, i, j)
|
rescale_ee_deriv_e(ii, i, j) = elec_dist_deriv_e(ii, i, j)
|
||||||
end do
|
end do
|
||||||
@ -48,7 +54,7 @@ BEGIN_PROVIDER [ double precision, rescale_ee_deriv_e, (4, nelec, nelec) ]
|
|||||||
(-kappa * rescale_ee_deriv_e(3, i, j) * rescale_ee_deriv_e(3, i, j))
|
(-kappa * rescale_ee_deriv_e(3, i, j) * rescale_ee_deriv_e(3, i, j))
|
||||||
do ii = 1, 4
|
do ii = 1, 4
|
||||||
rescale_ee_deriv_e(ii, i, j) = rescale_ee_deriv_e(ii, i, j) &
|
rescale_ee_deriv_e(ii, i, j) = rescale_ee_deriv_e(ii, i, j) &
|
||||||
* dexp(-kappa * elec_dist(i, j))
|
* f
|
||||||
enddo
|
enddo
|
||||||
enddo
|
enddo
|
||||||
enddo
|
enddo
|
||||||
@ -76,9 +82,11 @@ BEGIN_PROVIDER [ double precision, rescale_en_deriv_e, (4, nelec, nnuc) ]
|
|||||||
! Dimension 4 : d2x + d2y + d2z
|
! Dimension 4 : d2x + d2y + d2z
|
||||||
END_DOC
|
END_DOC
|
||||||
integer :: i, ii, a
|
integer :: i, ii, a
|
||||||
|
double precision :: f
|
||||||
|
|
||||||
do a = 1, nnuc
|
do a = 1, nnuc
|
||||||
do i = 1, nelec
|
do i = 1, nelec
|
||||||
|
f = 1.d0 - kappa*rescale_en(i,a) ! == dexp(-kappa * elnuc_dist(i, a))
|
||||||
do ii = 1, 4
|
do ii = 1, 4
|
||||||
rescale_en_deriv_e(ii, i, a) = elnuc_dist_deriv_e(ii, i, a)
|
rescale_en_deriv_e(ii, i, a) = elnuc_dist_deriv_e(ii, i, a)
|
||||||
end do
|
end do
|
||||||
@ -88,7 +96,7 @@ BEGIN_PROVIDER [ double precision, rescale_en_deriv_e, (4, nelec, nnuc) ]
|
|||||||
(-kappa * rescale_en_deriv_e(3, i, a) * rescale_en_deriv_e(3, i, a))
|
(-kappa * rescale_en_deriv_e(3, i, a) * rescale_en_deriv_e(3, i, a))
|
||||||
do ii = 1, 4
|
do ii = 1, 4
|
||||||
rescale_en_deriv_e(ii, i, a) = rescale_en_deriv_e(ii, i, a) &
|
rescale_en_deriv_e(ii, i, a) = rescale_en_deriv_e(ii, i, a) &
|
||||||
* dexp(-kappa * elnuc_dist(i, a))
|
* f
|
||||||
enddo
|
enddo
|
||||||
enddo
|
enddo
|
||||||
enddo
|
enddo
|
||||||
@ -100,18 +108,28 @@ BEGIN_PROVIDER [double precision, rescale_een_e, (nelec, nelec, 0:ncord)]
|
|||||||
! R = exp(-kappa r) for electron-electron for $J_{een}$
|
! R = exp(-kappa r) for electron-electron for $J_{een}$
|
||||||
END_DOC
|
END_DOC
|
||||||
integer :: i, j, l
|
integer :: i, j, l
|
||||||
double precision :: kappa_l
|
double precision :: x
|
||||||
|
double precision, parameter :: f = dexp(1.d0)
|
||||||
|
|
||||||
do l = 0, ncord
|
rescale_een_e(:, :, 0) = 1.d0
|
||||||
kappa_l = -dble(l) * kappa
|
|
||||||
do j = 1, nelec
|
do j = 1, nelec
|
||||||
do i = 1, nelec
|
do i = 1, j-1
|
||||||
rescale_een_e(i, j, l) = kappa_l * elec_dist(i, j)
|
x = dexp(-kappa * elec_dist(i, j))
|
||||||
enddo
|
rescale_een_e(i, j, 1) = x
|
||||||
|
rescale_een_e(j, i, 1) = x
|
||||||
enddo
|
enddo
|
||||||
enddo
|
enddo
|
||||||
|
|
||||||
rescale_een_e = dexp(rescale_een_e)
|
do l = 2, ncord
|
||||||
|
do j = 1, nelec
|
||||||
|
do i = 1, j-1
|
||||||
|
x = rescale_een_e(i, j, l-1) * rescale_een_e(i, j, 1)
|
||||||
|
rescale_een_e(i, j, l) = x
|
||||||
|
rescale_een_e(j, i, l) = x
|
||||||
|
enddo
|
||||||
|
enddo
|
||||||
|
enddo
|
||||||
|
|
||||||
do l = 0, ncord
|
do l = 0, ncord
|
||||||
do j = 1, nelec
|
do j = 1, nelec
|
||||||
@ -126,18 +144,24 @@ BEGIN_PROVIDER [double precision, rescale_een_n, (nelec, nnuc, 0:ncord)]
|
|||||||
! R = exp(-kappa r) for electron-electron for $J_{een}$
|
! R = exp(-kappa r) for electron-electron for $J_{een}$
|
||||||
END_DOC
|
END_DOC
|
||||||
integer :: i, a, l
|
integer :: i, a, l
|
||||||
double precision :: kappa_l
|
double precision :: x
|
||||||
|
double precision, parameter :: f = dexp(1.d0)
|
||||||
|
|
||||||
do l = 0, ncord
|
rescale_een_n(:,:,0) = 1.d0
|
||||||
kappa_l = - dble(l) * kappa
|
|
||||||
do a = 1, nnuc
|
do a = 1, nnuc
|
||||||
do i = 1, nelec
|
do i = 1, nelec
|
||||||
rescale_een_n(i, a, l) = kappa_l * elnuc_dist(i, a)
|
rescale_een_n(i, a, 1) = dexp(-kappa * elnuc_dist(i, a))
|
||||||
enddo
|
|
||||||
enddo
|
enddo
|
||||||
enddo
|
enddo
|
||||||
|
|
||||||
rescale_een_n = dexp(rescale_een_n)
|
do l = 2, ncord
|
||||||
|
do a = 1, nnuc
|
||||||
|
do i = 1, nelec
|
||||||
|
rescale_een_n(i, a, l) = rescale_een_n(i, a, l-1) * rescale_een_n(i, a, 1)
|
||||||
|
enddo
|
||||||
|
enddo
|
||||||
|
enddo
|
||||||
|
|
||||||
END_PROVIDER
|
END_PROVIDER
|
||||||
|
|
||||||
@ -186,7 +210,7 @@ BEGIN_PROVIDER [double precision, elnuc_dist_deriv_e, (4, nelec, nnuc)]
|
|||||||
integer :: i, ii, a
|
integer :: i, ii, a
|
||||||
double precision :: ria_inv
|
double precision :: ria_inv
|
||||||
|
|
||||||
do a = 1, nnuc
|
do a = 1, nnuc
|
||||||
do i = 1, nelec
|
do i = 1, nelec
|
||||||
ria_inv = 1.0d0 / elnuc_dist(i, a)
|
ria_inv = 1.0d0 / elnuc_dist(i, a)
|
||||||
do ii = 1, 3
|
do ii = 1, 3
|
||||||
@ -207,6 +231,7 @@ BEGIN_PROVIDER [double precision, rescale_een_e_deriv_e, (4, nelec, nelec, 0:nco
|
|||||||
integer :: i, ii, j, l
|
integer :: i, ii, j, l
|
||||||
double precision :: kappa_l
|
double precision :: kappa_l
|
||||||
|
|
||||||
|
!TODO: Check if rescale_een_e_deriv_e(:,:,0) = 0.d0
|
||||||
do l = 0, ncord
|
do l = 0, ncord
|
||||||
kappa_l = - dble(l) * kappa
|
kappa_l = - dble(l) * kappa
|
||||||
do j = 1, nelec
|
do j = 1, nelec
|
||||||
@ -232,7 +257,7 @@ BEGIN_PROVIDER [double precision, rescale_een_e_deriv_e, (4, nelec, nelec, 0:nco
|
|||||||
enddo
|
enddo
|
||||||
enddo
|
enddo
|
||||||
END_PROVIDER
|
END_PROVIDER
|
||||||
|
|
||||||
BEGIN_PROVIDER [double precision, elec_dist_deriv_e, (4, nelec, nelec)]
|
BEGIN_PROVIDER [double precision, elec_dist_deriv_e, (4, nelec, nelec)]
|
||||||
BEGIN_DOC
|
BEGIN_DOC
|
||||||
! Derivative of R_{ij} wrt x
|
! Derivative of R_{ij} wrt x
|
||||||
|
Loading…
Reference in New Issue
Block a user