9
1
mirror of https://github.com/QuantumPackage/qp2.git synced 2024-12-30 15:15:38 +01:00

added Charge_Harmonizer for numerical integrals

This commit is contained in:
AbdAmmar 2024-02-01 08:57:07 +01:00
parent cc334b34b7
commit 9e1b2f35d3
2 changed files with 245 additions and 59 deletions

View File

@ -98,14 +98,20 @@ BEGIN_PROVIDER [double precision, int2_grad1_u2e_ao, (ao_num, ao_num, n_points_f
END_DOC END_DOC
implicit none implicit none
integer :: ipoint, i, j, m, jpoint integer :: ipoint, i, j, m, jpoint
double precision :: time0, time1 integer :: n_blocks, n_rest, n_pass
double precision :: x, y, z, r2 integer :: i_blocks, i_rest, i_pass, ii
double precision :: dx, dy, dz double precision :: mem, n_double
double precision :: tmp_ct double precision :: time0, time1
double precision :: tmp0, tmp1, tmp2 double precision :: x, y, z, r2
double precision :: tmp0_x, tmp0_y, tmp0_z double precision :: dx, dy, dz
double precision :: tmp1_x, tmp1_y, tmp1_z double precision :: tmp_ct
double precision :: tmp0, tmp1, tmp2
double precision :: tmp0_x, tmp0_y, tmp0_z
double precision :: tmp1_x, tmp1_y, tmp1_z
double precision, allocatable :: tmp(:,:,:)
double precision, allocatable :: tmp_grad1_u12(:,:,:)
PROVIDE j2e_type PROVIDE j2e_type
PROVIDE Env_type PROVIDE Env_type
@ -113,70 +119,172 @@ BEGIN_PROVIDER [double precision, int2_grad1_u2e_ao, (ao_num, ao_num, n_points_f
call wall_time(time0) call wall_time(time0)
print*, ' providing int2_grad1_u2e_ao ...' print*, ' providing int2_grad1_u2e_ao ...'
if( (j2e_type .eq. "Mu") .and. & if(tc_integ_type .eq. "numeric") then
( (env_type .eq. "None") .or. (env_type .eq. "Prod_Gauss") .or. (env_type .eq. "Sum_Gauss") ) ) then
PROVIDE mu_erf PROVIDE final_weight_at_r_vector_extra aos_in_r_array_extra
PROVIDE env_type env_val env_grad
PROVIDE Ir2_Mu_long_Du_0 Ir2_Mu_long_Du_x Ir2_Mu_long_Du_y Ir2_Mu_long_Du_z Ir2_Mu_long_Du_2
PROVIDE Ir2_Mu_gauss_Du
tmp_ct = 0.5d0 / (dsqrt(dacos(-1.d0)) * mu_erf) allocate(tmp(n_points_extra_final_grid,ao_num,ao_num))
!$OMP PARALLEL &
!$OMP PARALLEL & !$OMP DEFAULT (NONE) &
!$OMP DEFAULT (NONE) & !$OMP PRIVATE (j, i, jpoint) &
!$OMP PRIVATE (ipoint, i, j, x, y, z, r2, dx, dy, dz, tmp1, tmp2, & !$OMP SHARED (tmp, ao_num, n_points_extra_final_grid, final_weight_at_r_vector_extra, aos_in_r_array_extra_transp)
!$OMP tmp0_x, tmp0_y, tmp0_z, tmp1_x, tmp1_y, tmp1_z) &
!$OMP SHARED (ao_num, n_points_final_grid, final_grid_points, &
!$OMP tmp_ct, env_val, env_grad, Ir2_Mu_long_Du_0, &
!$OMP Ir2_Mu_long_Du_x, Ir2_Mu_long_Du_y, &
!$OMP Ir2_Mu_long_Du_z, Ir2_Mu_gauss_Du, &
!$OMP Ir2_Mu_long_Du_2, int2_grad1_u2e_ao)
!$OMP DO SCHEDULE (static) !$OMP DO SCHEDULE (static)
do ipoint = 1, n_points_final_grid do j = 1, ao_num
do i = 1, ao_num
x = final_grid_points(1,ipoint) do jpoint = 1, n_points_extra_final_grid
y = final_grid_points(2,ipoint) tmp(jpoint,i,j) = final_weight_at_r_vector_extra(jpoint) * aos_in_r_array_extra_transp(jpoint,i) * aos_in_r_array_extra_transp(jpoint,j)
z = final_grid_points(3,ipoint)
r2 = x*x + y*y + z*z
dx = env_grad(1,ipoint)
dy = env_grad(2,ipoint)
dz = env_grad(3,ipoint)
tmp0_x = 0.5d0 * (env_val(ipoint) * x + r2 * dx)
tmp0_y = 0.5d0 * (env_val(ipoint) * y + r2 * dy)
tmp0_z = 0.5d0 * (env_val(ipoint) * z + r2 * dz)
tmp1 = 0.5d0 * env_val(ipoint)
tmp1_x = tmp_ct * dx
tmp1_y = tmp_ct * dy
tmp1_z = tmp_ct * dz
do j = 1, ao_num
do i = 1, ao_num
tmp2 = 0.5d0 * Ir2_Mu_long_Du_2(i,j,ipoint) - x * Ir2_Mu_long_Du_x(i,j,ipoint) - y * Ir2_Mu_long_Du_y(i,j,ipoint) - z * Ir2_Mu_long_Du_z(i,j,ipoint)
int2_grad1_u2e_ao(i,j,ipoint,1) = Ir2_Mu_long_Du_0(i,j,ipoint) * tmp0_x - tmp1 * Ir2_Mu_long_Du_x(i,j,ipoint) + dx * tmp2 - tmp1_x * Ir2_Mu_gauss_Du(i,j,ipoint)
int2_grad1_u2e_ao(i,j,ipoint,2) = Ir2_Mu_long_Du_0(i,j,ipoint) * tmp0_y - tmp1 * Ir2_Mu_long_Du_y(i,j,ipoint) + dy * tmp2 - tmp1_y * Ir2_Mu_gauss_Du(i,j,ipoint)
int2_grad1_u2e_ao(i,j,ipoint,3) = Ir2_Mu_long_Du_0(i,j,ipoint) * tmp0_z - tmp1 * Ir2_Mu_long_Du_z(i,j,ipoint) + dz * tmp2 - tmp1_z * Ir2_Mu_gauss_Du(i,j,ipoint)
enddo enddo
enddo enddo
enddo enddo
!$OMP END DO !$OMP END DO
!$OMP END PARALLEL !$OMP END PARALLEL
FREE Ir2_Mu_long_Du_0 Ir2_Mu_long_Du_x Ir2_Mu_long_Du_y Ir2_Mu_long_Du_z Ir2_Mu_long_Du_2 ! n_points_final_grid = n_blocks * n_pass + n_rest
FREE Ir2_Mu_gauss_Du call total_memory(mem)
mem = max(1.d0, qp_max_mem - mem)
n_double = mem * 1.d8
n_blocks = int(min(n_double / (n_points_extra_final_grid * 4.d0), 1.d0*n_points_final_grid))
n_rest = int(mod(n_points_final_grid, n_blocks))
n_pass = int((n_points_final_grid - n_rest) / n_blocks)
call write_int(6, n_pass, 'Number of passes')
call write_int(6, n_blocks, 'Size of the blocks')
call write_int(6, n_rest, 'Size of the last block')
allocate(tmp_grad1_u12(n_points_extra_final_grid,n_blocks,3))
do i_pass = 1, n_pass
ii = (i_pass-1)*n_blocks + 1
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i_blocks, ipoint) &
!$OMP SHARED (n_blocks, n_points_extra_final_grid, ii, &
!$OMP final_grid_points, tmp_grad1_u12)
!$OMP DO
do i_blocks = 1, n_blocks
ipoint = ii - 1 + i_blocks ! r1
call get_grad1_u12_2e_r1_seq(ipoint, n_points_extra_final_grid, tmp_grad1_u12(1,i_blocks,1) &
, tmp_grad1_u12(1,i_blocks,2) &
, tmp_grad1_u12(1,i_blocks,3))
enddo
!$OMP END DO
!$OMP END PARALLEL
do m = 1, 3
call dgemm( "T", "N", ao_num*ao_num, n_blocks, n_points_extra_final_grid, 1.d0 &
, tmp(1,1,1), n_points_extra_final_grid, tmp_grad1_u12(1,1,m), n_points_extra_final_grid &
, 0.d0, int2_grad1_u2e_ao(1,1,ii,m), ao_num*ao_num)
enddo
enddo
deallocate(tmp_grad1_u12)
if(n_rest .gt. 0) then
allocate(tmp_grad1_u12(n_points_extra_final_grid,n_rest,3))
ii = n_pass*n_blocks + 1
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i_rest, ipoint) &
!$OMP SHARED (n_rest, n_points_extra_final_grid, ii, &
!$OMP final_grid_points, tmp_grad1_u12)
!$OMP DO
do i_rest = 1, n_rest
ipoint = ii - 1 + i_rest ! r1
call get_grad1_u12_2e_r1_seq(ipoint, n_points_extra_final_grid, tmp_grad1_u12(1,i_rest,1) &
, tmp_grad1_u12(1,i_rest,2) &
, tmp_grad1_u12(1,i_rest,3))
enddo
!$OMP END DO
!$OMP END PARALLEL
do m = 1, 3
call dgemm( "T", "N", ao_num*ao_num, n_rest, n_points_extra_final_grid, 1.d0 &
, tmp(1,1,1), n_points_extra_final_grid, tmp_grad1_u12(1,1,m), n_points_extra_final_grid &
, 0.d0, int2_grad1_u2e_ao(1,1,ii,m), ao_num*ao_num)
enddo
deallocate(tmp_grad1_u12)
endif
deallocate(tmp)
elseif(tc_integ_type .eq. "semi-analytic") then
if( (j2e_type .eq. "Mu") .and. &
( (env_type .eq. "None") .or. (env_type .eq. "Prod_Gauss") .or. (env_type .eq. "Sum_Gauss") ) ) then
PROVIDE mu_erf
PROVIDE env_type env_val env_grad
PROVIDE Ir2_Mu_long_Du_0 Ir2_Mu_long_Du_x Ir2_Mu_long_Du_y Ir2_Mu_long_Du_z Ir2_Mu_long_Du_2
PROVIDE Ir2_Mu_gauss_Du
tmp_ct = 0.5d0 / (dsqrt(dacos(-1.d0)) * mu_erf)
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (ipoint, i, j, x, y, z, r2, dx, dy, dz, tmp1, tmp2, &
!$OMP tmp0_x, tmp0_y, tmp0_z, tmp1_x, tmp1_y, tmp1_z) &
!$OMP SHARED (ao_num, n_points_final_grid, final_grid_points, &
!$OMP tmp_ct, env_val, env_grad, Ir2_Mu_long_Du_0, &
!$OMP Ir2_Mu_long_Du_x, Ir2_Mu_long_Du_y, &
!$OMP Ir2_Mu_long_Du_z, Ir2_Mu_gauss_Du, &
!$OMP Ir2_Mu_long_Du_2, int2_grad1_u2e_ao)
!$OMP DO SCHEDULE (static)
do ipoint = 1, n_points_final_grid
x = final_grid_points(1,ipoint)
y = final_grid_points(2,ipoint)
z = final_grid_points(3,ipoint)
r2 = x*x + y*y + z*z
dx = env_grad(1,ipoint)
dy = env_grad(2,ipoint)
dz = env_grad(3,ipoint)
tmp0_x = 0.5d0 * (env_val(ipoint) * x + r2 * dx)
tmp0_y = 0.5d0 * (env_val(ipoint) * y + r2 * dy)
tmp0_z = 0.5d0 * (env_val(ipoint) * z + r2 * dz)
tmp1 = 0.5d0 * env_val(ipoint)
tmp1_x = tmp_ct * dx
tmp1_y = tmp_ct * dy
tmp1_z = tmp_ct * dz
do j = 1, ao_num
do i = 1, ao_num
tmp2 = 0.5d0 * Ir2_Mu_long_Du_2(i,j,ipoint) - x * Ir2_Mu_long_Du_x(i,j,ipoint) - y * Ir2_Mu_long_Du_y(i,j,ipoint) - z * Ir2_Mu_long_Du_z(i,j,ipoint)
int2_grad1_u2e_ao(i,j,ipoint,1) = Ir2_Mu_long_Du_0(i,j,ipoint) * tmp0_x - tmp1 * Ir2_Mu_long_Du_x(i,j,ipoint) + dx * tmp2 - tmp1_x * Ir2_Mu_gauss_Du(i,j,ipoint)
int2_grad1_u2e_ao(i,j,ipoint,2) = Ir2_Mu_long_Du_0(i,j,ipoint) * tmp0_y - tmp1 * Ir2_Mu_long_Du_y(i,j,ipoint) + dy * tmp2 - tmp1_y * Ir2_Mu_gauss_Du(i,j,ipoint)
int2_grad1_u2e_ao(i,j,ipoint,3) = Ir2_Mu_long_Du_0(i,j,ipoint) * tmp0_z - tmp1 * Ir2_Mu_long_Du_z(i,j,ipoint) + dz * tmp2 - tmp1_z * Ir2_Mu_gauss_Du(i,j,ipoint)
enddo
enddo
enddo
!$OMP END DO
!$OMP END PARALLEL
FREE Ir2_Mu_long_Du_0 Ir2_Mu_long_Du_x Ir2_Mu_long_Du_y Ir2_Mu_long_Du_z Ir2_Mu_long_Du_2
FREE Ir2_Mu_gauss_Du
else
print *, ' Error in int2_grad1_u2e_ao: Unknown Jastrow'
stop
endif ! j2e_type
else else
print *, ' Error in int2_grad1_u2e_ao: Unknown Jastrow' print *, ' Error in int2_grad1_u2e_ao: Unknown tc_integ_type'
stop stop
endif ! j2e_type endif ! tc_integ_type
call wall_time(time1) call wall_time(time1)
print*, ' wall time for int2_grad1_u2e_ao (min) =', (time1-time0)/60.d0 print*, ' wall time for int2_grad1_u2e_ao (min) =', (time1-time0)/60.d0

View File

@ -395,3 +395,81 @@ end
! --- ! ---
subroutine get_grad1_u12_2e_r1_seq(ipoint, n_grid2, resx, resy, resz)
BEGIN_DOC
!
! grad_1 u_2e(r1,r2)
!
! we use grid for r1 and extra_grid for r2
!
END_DOC
implicit none
integer, intent(in) :: ipoint, n_grid2
double precision, intent(out) :: resx(n_grid2), resy(n_grid2), resz(n_grid2)
integer :: jpoint
double precision :: env_r1, tmp
double precision :: grad1_env(3), r1(3)
double precision, allocatable :: env_r2(:)
double precision, allocatable :: u2b_r12(:)
double precision, allocatable :: gradx1_u2b(:), grady1_u2b(:), gradz1_u2b(:)
double precision, external :: env_nucl
PROVIDE j1e_type j2e_type env_type
PROVIDE final_grid_points
PROVIDE final_grid_points_extra
r1(1) = final_grid_points(1,ipoint)
r1(2) = final_grid_points(2,ipoint)
r1(3) = final_grid_points(3,ipoint)
if( (j2e_type .eq. "Mu") .or. &
(j2e_type .eq. "Mur") .or. &
(j2e_type .eq. "Boys") ) then
if(env_type .eq. "None") then
call grad1_j12_r1_seq(r1, n_grid2, resx, resy, resz)
else
! u(r1,r2) = j12_mu(r12) x v(r1) x v(r2)
! grad1 u(r1, r2) = [(grad1 j12_mu) v(r1) + j12_mu grad1 v(r1)] v(r2)
allocate(env_r2(n_grid2))
allocate(u2b_r12(n_grid2))
allocate(gradx1_u2b(n_grid2))
allocate(grady1_u2b(n_grid2))
allocate(gradz1_u2b(n_grid2))
env_r1 = env_nucl(r1)
call grad1_env_nucl(r1, grad1_env)
call env_nucl_r1_seq(n_grid2, env_r2)
call j12_r1_seq(r1, n_grid2, u2b_r12)
call grad1_j12_r1_seq(r1, n_grid2, gradx1_u2b, grady1_u2b, gradz1_u2b)
do jpoint = 1, n_points_extra_final_grid
resx(jpoint) = (gradx1_u2b(jpoint) * env_r1 + u2b_r12(jpoint) * grad1_env(1)) * env_r2(jpoint)
resy(jpoint) = (grady1_u2b(jpoint) * env_r1 + u2b_r12(jpoint) * grad1_env(2)) * env_r2(jpoint)
resz(jpoint) = (gradz1_u2b(jpoint) * env_r1 + u2b_r12(jpoint) * grad1_env(3)) * env_r2(jpoint)
enddo
deallocate(env_r2, u2b_r12, gradx1_u2b, grady1_u2b, gradz1_u2b)
endif ! env_type
else
print *, ' Error in get_grad1_u12_withsq_r1_seq: Unknown Jastrow'
stop
endif ! j2e_type
return
end
! ---