LCPQ/qp2
9
1
Fork 0
mirror of https://github.com/QuantumPackage/qp2.git synced 2024-12-21 11:03:29 +01:00
Code Releases Activity

Merge branch 'dev-stable-tc-scf' of https://github.com/QuantumPackage/qp2 into dev-stable-tc-scf

Browse Source
This commit is contained in:
Abdallah Ammar 2023-04-16 19:27:22 +02:00
commit eea6758d26
46 changed files with 3371 additions and 694 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
external/qp2-dependencies vendored

@ -1 +1 @@
Subproject commit ce14f57b50511825a9fedb096749200779d3f4d4
Subproject commit 6e23ebac001acae91d1c762ca934e09a9b7d614a

43
scripts/Hn.py Normal file
View File

@ -0,0 +1,43 @@
#!/usr/bin/env python
import sys
from math import *
arg = sys.argv
#f = open('data_dft','r')
n = int(sys.argv[1])
r = float(sys.argv[2])
f = open('H'+str(n)+'_'+str(r)+'.xyz','w')
string=str(n)+"\n"
f.write(string)
string="\n"
f.write(string)
for i in range(n):
x = r * cos(2.* i* pi/n)
y = r * sin(2.* i* pi/n)
z = 0.
string="H "+str(x)+" "+str(y)+" "+str(z)+"\n"
f.write(string)
#lines = f.readlines()
#cipsi_dft= []
#
#dissoc = []
#dissoc.append(float(-76.0179223470363))
#dissoc.append(float(-76.0592367866993))
#dissoc.append(float(-76.0678739715659))
#delta_e = []
#
#for line in lines:
# data = line.split()
# if(len(data)>0):
# dft=float(data[1])
# fci=float(data[2])
# e=fci+dft
# cipsi_dft.append(e)
#
#print(*cipsi_dft,sep=" & ")
#
#for i in 0,1,2:
# delta_e.append(1000.*(dissoc[i] - cipsi_dft[i]))
#
#print(*delta_e,sep=" & ")
#

2
scripts/get_fci_tc_conv.sh Executable file
View File

@ -0,0 +1,2 @@
file=$1
grep "Ndet,E,E+PT2,E+RPT2,|PT2|=" $file | cut -d "=" -f 2 > ${file}.conv_fci_tc

48
src/ao_many_one_e_ints/grad2_jmu_manu.irp.f
View File

@ -38,7 +38,7 @@ BEGIN_PROVIDER [ double precision, int2_grad1u2_grad2u2_j1b2_test, (ao_num, ao_n
!$OMP expo_gauss_1_erf_x_2, coef_gauss_1_erf_x_2, &
!$OMP List_comb_thr_b3_coef, List_comb_thr_b3_expo, &
!$OMP List_comb_thr_b3_cent, int2_grad1u2_grad2u2_j1b2_test, ao_abs_comb_b3_j1b, &
!$OMP ao_overlap_abs,sq_pi_3_2)
!$OMP ao_overlap_abs,sq_pi_3_2,thrsh_cycle_tc)
!$OMP DO SCHEDULE(dynamic)
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
@ -46,7 +46,7 @@ BEGIN_PROVIDER [ double precision, int2_grad1u2_grad2u2_j1b2_test, (ao_num, ao_n
r(3) = final_grid_points(3,ipoint)
do i = 1, ao_num
do j = i, ao_num
if(ao_overlap_abs(j,i) .lt. 1.d-12) then
if(ao_overlap_abs(j,i) .lt. thrsh_cycle_tc) then
cycle
endif
@ -58,7 +58,7 @@ BEGIN_PROVIDER [ double precision, int2_grad1u2_grad2u2_j1b2_test, (ao_num, ao_n
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_1_erf_x_2(i_fit)
coef_fit = -0.25d0 * coef_gauss_1_erf_x_2(i_fit)
if(dabs(coef_fit*int_j1b*sq_pi_3_2*(expo_fit)**(-1.5d0)).lt.1.d-10)cycle
! if(dabs(coef_fit*int_j1b*sq_pi_3_2*(expo_fit)**(-1.5d0)).lt.thrsh_cycle_tc)cycle
int_gauss = overlap_gauss_r12_ao(r, expo_fit, i, j)
int2_grad1u2_grad2u2_j1b2_test(j,i,ipoint) += coef_fit * int_gauss
enddo
@ -81,8 +81,7 @@ BEGIN_PROVIDER [ double precision, int2_grad1u2_grad2u2_j1b2_test, (ao_num, ao_n
!DIR$ FORCEINLINE
call gaussian_product(expo_fit,r,beta,B_center,factor_ij_1s,beta_ij,center_ij_1s)
coef_fit = -0.25d0 * coef_gauss_1_erf_x_2(i_fit) * coef
! if(dabs(coef_fit*factor_ij_1s*int_j1b).lt.1.d-10)cycle ! old version
if(dabs(coef_fit*factor_ij_1s*int_j1b*sq_pi_3_2*(beta_ij)**(-1.5d0)).lt.1.d-10)cycle
! if(dabs(coef_fit*factor_ij_1s*int_j1b*sq_pi_3_2*(beta_ij)**(-1.5d0)).lt.thrsh_cycle_tc)cycle
! call overlap_gauss_r12_ao_with1s_v(B_center, beta, final_grid_points_transp, &
! expo_fit, i, j, int_fit_v, n_points_final_grid)
int_gauss = overlap_gauss_r12_ao_with1s(B_center, beta, r, expo_fit, i, j)
@ -145,14 +144,14 @@ BEGIN_PROVIDER [ double precision, int2_grad1u2_grad2u2_j1b2_test_v, (ao_num, ao
!$OMP expo_gauss_1_erf_x_2, coef_gauss_1_erf_x_2, &
!$OMP List_comb_thr_b3_coef, List_comb_thr_b3_expo, &
!$OMP List_comb_thr_b3_cent, big_array,&
!$OMP ao_abs_comb_b3_j1b,ao_overlap_abs)
!$OMP ao_abs_comb_b3_j1b,ao_overlap_abs,thrsh_cycle_tc)
!
allocate(int_fit_v(n_points_final_grid))
!$OMP DO SCHEDULE(dynamic)
do i = 1, ao_num
do j = i, ao_num
if(ao_overlap_abs(j,i) .lt. 1.d-12) then
if(ao_overlap_abs(j,i) .lt. thrsh_cycle_tc) then
cycle
endif
@ -161,7 +160,6 @@ BEGIN_PROVIDER [ double precision, int2_grad1u2_grad2u2_j1b2_test_v, (ao_num, ao
coef = List_comb_thr_b3_coef (i_1s,j,i)
beta = List_comb_thr_b3_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b3_j1b(i_1s,j,i)
! if(dabs(coef)*dabs(int_j1b).lt.1.d-15)cycle
B_center(1) = List_comb_thr_b3_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b3_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b3_cent(3,i_1s,j,i)
@ -243,7 +241,7 @@ BEGIN_PROVIDER [ double precision, int2_u2_j1b2_test, (ao_num, ao_num, n_points_
!$OMP final_grid_points, ng_fit_jast, &
!$OMP expo_gauss_j_mu_x_2, coef_gauss_j_mu_x_2, &
!$OMP List_comb_thr_b3_coef, List_comb_thr_b3_expo,sq_pi_3_2, &
!$OMP List_comb_thr_b3_cent, int2_u2_j1b2_test,ao_abs_comb_b3_j1b)
!$OMP List_comb_thr_b3_cent, int2_u2_j1b2_test,ao_abs_comb_b3_j1b,thrsh_cycle_tc)
!$OMP DO
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
@ -260,11 +258,11 @@ BEGIN_PROVIDER [ double precision, int2_u2_j1b2_test, (ao_num, ao_num, n_points_
! --- --- ---
int_j1b = ao_abs_comb_b3_j1b(1,j,i)
if(dabs(int_j1b).lt.1.d-10) cycle
if(dabs(int_j1b).lt.thrsh_cycle_tc) cycle
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_j_mu_x_2(i_fit)
coef_fit = coef_gauss_j_mu_x_2(i_fit)
if(dabs(coef_fit*int_j1b*sq_pi_3_2*(expo_fit)**(-1.5d0)).lt.1.d-10)cycle
! if(dabs(coef_fit*int_j1b*sq_pi_3_2*(expo_fit)**(-1.5d0)).lt.thrsh_cycle_tc)cycle
int_fit = overlap_gauss_r12_ao(r, expo_fit, i, j)
tmp += coef_fit * int_fit
enddo
@ -278,7 +276,7 @@ BEGIN_PROVIDER [ double precision, int2_u2_j1b2_test, (ao_num, ao_num, n_points_
coef = List_comb_thr_b3_coef (i_1s,j,i)
beta = List_comb_thr_b3_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b3_j1b(i_1s,j,i)
if(dabs(coef)*dabs(int_j1b).lt.1.d-10)cycle
! if(dabs(coef)*dabs(int_j1b).lt.thrsh_cycle_tc)cycle
B_center(1) = List_comb_thr_b3_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b3_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b3_cent(3,i_1s,j,i)
@ -288,8 +286,7 @@ BEGIN_PROVIDER [ double precision, int2_u2_j1b2_test, (ao_num, ao_num, n_points_
coef_fit = coef_gauss_j_mu_x_2(i_fit)
!DIR$ FORCEINLINE
call gaussian_product(expo_fit,r,beta,B_center,factor_ij_1s,beta_ij,center_ij_1s)
! if(dabs(coef_fit*coef*factor_ij_1s*int_j1b).lt.1.d-10)cycle ! old version
if(dabs(coef_fit*coef*factor_ij_1s*int_j1b*sq_pi_3_2*(beta_ij)**(-1.5d0)).lt.1.d-10)cycle
! if(dabs(coef_fit*coef*factor_ij_1s*int_j1b*sq_pi_3_2*(beta_ij)**(-1.5d0)).lt.thrsh_cycle_tc)cycle
int_fit = overlap_gauss_r12_ao_with1s(B_center, beta, r, expo_fit, i, j)
tmp += coef * coef_fit * int_fit
enddo
@ -350,7 +347,7 @@ BEGIN_PROVIDER [ double precision, int2_u_grad1u_x_j1b2_test, (ao_num, ao_num, n
!$OMP final_grid_points, ng_fit_jast, &
!$OMP expo_gauss_j_mu_1_erf, coef_gauss_j_mu_1_erf, &
!$OMP List_comb_thr_b3_coef, List_comb_thr_b3_expo, &
!$OMP List_comb_thr_b3_cent, int2_u_grad1u_x_j1b2_test,ao_abs_comb_b3_j1b,sq_pi_3_2)
!$OMP List_comb_thr_b3_cent, int2_u_grad1u_x_j1b2_test,ao_abs_comb_b3_j1b,sq_pi_3_2,thrsh_cycle_tc)
!$OMP DO
do ipoint = 1, n_points_final_grid
@ -369,7 +366,7 @@ BEGIN_PROVIDER [ double precision, int2_u_grad1u_x_j1b2_test, (ao_num, ao_num, n
coef = List_comb_thr_b3_coef (i_1s,j,i)
beta = List_comb_thr_b3_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b3_j1b(i_1s,j,i)
if(dabs(coef)*dabs(int_j1b).lt.1.d-10)cycle
if(dabs(coef)*dabs(int_j1b).lt.thrsh_cycle_tc)cycle
B_center(1) = List_comb_thr_b3_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b3_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b3_cent(3,i_1s,j,i)
@ -392,8 +389,7 @@ BEGIN_PROVIDER [ double precision, int2_u_grad1u_x_j1b2_test, (ao_num, ao_num, n
expo_coef_1s = beta * expo_fit * alpha_1s_inv * dist
coef_tmp = coef * coef_fit * dexp(-expo_coef_1s)
sq_alpha = alpha_1s_inv * dsqrt(alpha_1s_inv)
! if(dabs(coef_tmp*int_j1b) .lt. 1d-10) cycle ! old version
if(dabs(coef_tmp*int_j1b*sq_pi_3_2*sq_alpha) .lt. 1d-10) cycle
! if(dabs(coef_tmp*int_j1b*sq_pi_3_2*sq_alpha) .lt. thrsh_cycle_tc) cycle
call NAI_pol_x_mult_erf_ao_with1s(i, j, alpha_1s, centr_1s, 1.d+9, r, int_fit)
@ -470,13 +466,13 @@ BEGIN_PROVIDER [ double precision, int2_u_grad1u_j1b2_test, (ao_num, ao_num, n_p
!$OMP expo_gauss_j_mu_1_erf, coef_gauss_j_mu_1_erf, &
!$OMP ao_prod_dist_grid, ao_prod_sigma, ao_overlap_abs_grid,ao_prod_center,dsqpi_3_2, &
!$OMP List_comb_thr_b3_coef, List_comb_thr_b3_expo, ao_abs_comb_b3_j1b, &
!$OMP List_comb_thr_b3_cent, int2_u_grad1u_j1b2_test)
!$OMP List_comb_thr_b3_cent, int2_u_grad1u_j1b2_test,thrsh_cycle_tc)
!$OMP DO
do ipoint = 1, n_points_final_grid
do i = 1, ao_num
do j = i, ao_num
if(dabs(ao_overlap_abs_grid(j,i)).lt.1.d-10) cycle
if(dabs(ao_overlap_abs_grid(j,i)).lt.thrsh_cycle_tc) cycle
r(1) = final_grid_points(1,ipoint)
r(2) = final_grid_points(2,ipoint)
@ -489,10 +485,10 @@ BEGIN_PROVIDER [ double precision, int2_u_grad1u_j1b2_test, (ao_num, ao_num, n_p
! --- --- ---
int_j1b = ao_abs_comb_b3_j1b(1,j,i)
if(dabs(int_j1b).lt.1.d-10) cycle
! if(dabs(int_j1b).lt.thrsh_cycle_tc) cycle
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_j_mu_1_erf(i_fit)
if(dabs(int_j1b)*dsqpi_3_2*expo_fit**(-1.5d0).lt.1.d-15) cycle
! if(dabs(int_j1b)*dsqpi_3_2*expo_fit**(-1.5d0).lt.thrsh_cycle_tc) cycle
coef_fit = coef_gauss_j_mu_1_erf(i_fit)
int_fit = NAI_pol_mult_erf_ao_with1s(i, j, expo_fit, r, 1.d+9, r)
tmp += coef_fit * int_fit
@ -507,7 +503,7 @@ BEGIN_PROVIDER [ double precision, int2_u_grad1u_j1b2_test, (ao_num, ao_num, n_p
coef = List_comb_thr_b3_coef (i_1s,j,i)
beta = List_comb_thr_b3_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b3_j1b(i_1s,j,i)
if(dabs(coef)*dabs(int_j1b).lt.1.d-10)cycle
! if(dabs(coef)*dabs(int_j1b).lt.thrsh_cycle_tc)cycle
B_center(1) = List_comb_thr_b3_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b3_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b3_cent(3,i_1s,j,i)
@ -517,7 +513,7 @@ BEGIN_PROVIDER [ double precision, int2_u_grad1u_j1b2_test, (ao_num, ao_num, n_p
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_j_mu_1_erf(i_fit)
call gaussian_product(expo_fit,r,beta,B_center,factor_ij_1s,beta_ij,center_ij_1s)
if(factor_ij_1s*dabs(coef*int_j1b)*dsqpi_3_2*beta_ij**(-1.5d0).lt.1.d-15)cycle
! if(factor_ij_1s*dabs(coef*int_j1b)*dsqpi_3_2*beta_ij**(-1.5d0).lt.thrsh_cycle_tc)cycle
coef_fit = coef_gauss_j_mu_1_erf(i_fit)
alpha_1s = beta + expo_fit
@ -527,9 +523,9 @@ BEGIN_PROVIDER [ double precision, int2_u_grad1u_j1b2_test, (ao_num, ao_num, n_p
centr_1s(3) = alpha_1s_inv * (beta * B_center(3) + expo_fit * r(3))
expo_coef_1s = beta * expo_fit * alpha_1s_inv * dist
if(expo_coef_1s .gt. 20.d0) cycle
! if(expo_coef_1s .gt. 20.d0) cycle
coef_tmp = coef * coef_fit * dexp(-expo_coef_1s)
if(dabs(coef_tmp) .lt. 1d-08) cycle
! if(dabs(coef_tmp) .lt. 1d-08) cycle
int_fit = NAI_pol_mult_erf_ao_with1s(i, j, alpha_1s, centr_1s, 1.d+9, r)

46
src/ao_many_one_e_ints/grad_lapl_jmu_manu.irp.f
View File

@ -31,7 +31,7 @@ BEGIN_PROVIDER [ double precision, v_ij_erf_rk_cst_mu_j1b_test, (ao_num, ao_num,
!$OMP SHARED (n_points_final_grid, ao_num, List_comb_thr_b2_size, final_grid_points, &
!$OMP List_comb_thr_b2_coef, List_comb_thr_b2_expo, List_comb_thr_b2_cent,ao_abs_comb_b2_j1b, &
!$OMP v_ij_erf_rk_cst_mu_j1b_test, mu_erf, &
!$OMP ao_overlap_abs_grid,ao_prod_center,ao_prod_sigma,dsqpi_3_2)
!$OMP ao_overlap_abs_grid,ao_prod_center,ao_prod_sigma,dsqpi_3_2,thrsh_cycle_tc)
!$OMP DO
!do ipoint = 1, 10
do ipoint = 1, n_points_final_grid
@ -41,7 +41,7 @@ BEGIN_PROVIDER [ double precision, v_ij_erf_rk_cst_mu_j1b_test, (ao_num, ao_num,
do i = 1, ao_num
do j = i, ao_num
if(dabs(ao_overlap_abs_grid(j,i)).lt.1.d-20)cycle
if(dabs(ao_overlap_abs_grid(j,i)).lt.thrsh_cycle_tc)cycle
tmp = 0.d0
do i_1s = 1, List_comb_thr_b2_size(j,i)
@ -49,7 +49,7 @@ BEGIN_PROVIDER [ double precision, v_ij_erf_rk_cst_mu_j1b_test, (ao_num, ao_num,
coef = List_comb_thr_b2_coef (i_1s,j,i)
beta = List_comb_thr_b2_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b2_j1b(i_1s,j,i)
if(dabs(coef)*dabs(int_j1b).lt.1.d-10)cycle
! if(dabs(coef)*dabs(int_j1b).lt.thrsh_cycle_tc)cycle
B_center(1) = List_comb_thr_b2_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b2_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b2_cent(3,i_1s,j,i)
@ -110,7 +110,7 @@ BEGIN_PROVIDER [ double precision, x_v_ij_erf_rk_cst_mu_j1b_test, (ao_num, ao_nu
!$OMP SHARED (n_points_final_grid, ao_num, List_comb_thr_b2_size, final_grid_points,&
!$OMP List_comb_thr_b2_coef, List_comb_thr_b2_expo, List_comb_thr_b2_cent, &
!$OMP x_v_ij_erf_rk_cst_mu_j1b_test, mu_erf,ao_abs_comb_b2_j1b, &
!$OMP ao_overlap_abs_grid,ao_prod_center,ao_prod_sigma)
!$OMP ao_overlap_abs_grid,ao_prod_center,ao_prod_sigma,thrsh_cycle_tc)
! !$OMP ao_overlap_abs_grid,ao_prod_center,ao_prod_sigma,dsqpi_3_2,expo_erfc_mu_gauss)
!$OMP DO
do ipoint = 1, n_points_final_grid
@ -120,7 +120,7 @@ BEGIN_PROVIDER [ double precision, x_v_ij_erf_rk_cst_mu_j1b_test, (ao_num, ao_nu
do i = 1, ao_num
do j = i, ao_num
if(dabs(ao_overlap_abs_grid(j,i)).lt.1.d-10)cycle
if(dabs(ao_overlap_abs_grid(j,i)).lt.thrsh_cycle_tc)cycle
tmp_x = 0.d0
tmp_y = 0.d0
@ -130,19 +130,11 @@ BEGIN_PROVIDER [ double precision, x_v_ij_erf_rk_cst_mu_j1b_test, (ao_num, ao_nu
coef = List_comb_thr_b2_coef (i_1s,j,i)
beta = List_comb_thr_b2_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b2_j1b(i_1s,j,i)
if(dabs(coef)*dabs(int_j1b).lt.1.d-10)cycle
! if(dabs(coef)*dabs(int_j1b).lt.thrsh_cycle_tc)cycle
B_center(1) = List_comb_thr_b2_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b2_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b2_cent(3,i_1s,j,i)
! if(ao_prod_center(1,j,i).ne.10000.d0)then
! ! approximate 1 - erf(mu r12) by a gaussian * 10
! !DIR$ FORCEINLINE
! call gaussian_product(expo_erfc_mu_gauss,r, &
! ao_prod_sigma(j,i),ao_prod_center(1,j,i), &
! factor_ij_1s,beta_ij,center_ij_1s)
! if(dabs(coef * factor_ij_1s*int_j1b*10.d0 * dsqpi_3_2 * beta_ij**(-1.5d0)).lt.1.d-10)cycle
! endif
call NAI_pol_x_mult_erf_ao_with1s(i, j, beta, B_center, mu_erf, r, ints )
call NAI_pol_x_mult_erf_ao_with1s(i, j, beta, B_center, 1.d+9, r, ints_coulomb)
@ -216,7 +208,7 @@ BEGIN_PROVIDER [ double precision, v_ij_u_cst_mu_j1b_test, (ao_num, ao_num, n_po
!$OMP expo_gauss_j_mu_x, coef_gauss_j_mu_x, &
!$OMP List_comb_thr_b2_coef, List_comb_thr_b2_expo,List_comb_thr_b2_size, &
!$OMP List_comb_thr_b2_cent, v_ij_u_cst_mu_j1b_test,ao_abs_comb_b2_j1b, &
!$OMP ao_overlap_abs_grid,ao_prod_center,ao_prod_sigma,dsqpi_3_2)
!$OMP ao_overlap_abs_grid,ao_prod_center,ao_prod_sigma,dsqpi_3_2,thrsh_cycle_tc)
!$OMP DO
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
@ -225,7 +217,7 @@ BEGIN_PROVIDER [ double precision, v_ij_u_cst_mu_j1b_test, (ao_num, ao_num, n_po
do i = 1, ao_num
do j = i, ao_num
if(dabs(ao_overlap_abs_grid(j,i)).lt.1.d-20)cycle
if(dabs(ao_overlap_abs_grid(j,i)).lt.thrsh_cycle_tc)cycle
tmp = 0.d0
@ -234,11 +226,11 @@ BEGIN_PROVIDER [ double precision, v_ij_u_cst_mu_j1b_test, (ao_num, ao_num, n_po
! --- --- ---
int_j1b = ao_abs_comb_b2_j1b(1,j,i)
if(dabs(int_j1b).lt.1.d-10) cycle
! if(dabs(int_j1b).lt.thrsh_cycle_tc) cycle
do i_fit = 1, ng_fit_jast
expo_fit = expo_gauss_j_mu_x(i_fit)
coef_fit = coef_gauss_j_mu_x(i_fit)
if(ao_overlap_abs_grid(j,i).lt.1.d-15) cycle
! if(ao_overlap_abs_grid(j,i).lt.thrsh_cycle_tc) cycle
int_fit = overlap_gauss_r12_ao(r, expo_fit, i, j)
tmp += coef_fit * int_fit
enddo
@ -251,7 +243,7 @@ BEGIN_PROVIDER [ double precision, v_ij_u_cst_mu_j1b_test, (ao_num, ao_num, n_po
coef = List_comb_thr_b2_coef (i_1s,j,i)
beta = List_comb_thr_b2_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b2_j1b(i_1s,j,i)
if(dabs(coef)*dabs(int_j1b).lt.1.d-10)cycle
! if(dabs(coef)*dabs(int_j1b).lt.thrsh_cycle_tc)cycle
B_center(1) = List_comb_thr_b2_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b2_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b2_cent(3,i_1s,j,i)
@ -259,9 +251,9 @@ BEGIN_PROVIDER [ double precision, v_ij_u_cst_mu_j1b_test, (ao_num, ao_num, n_po
expo_fit = expo_gauss_j_mu_x(i_fit)
coef_fit = coef_gauss_j_mu_x(i_fit)
coeftot = coef * coef_fit
if(dabs(coeftot).lt.1.d-15)cycle
! if(dabs(coeftot).lt.thrsh_cycle_tc)cycle
call gaussian_product(beta,B_center,expo_fit,r,factor_ij_1s_u,beta_ij_u,center_ij_1s_u)
if(factor_ij_1s_u*ao_overlap_abs_grid(j,i).lt.1.d-15)cycle
! if(factor_ij_1s_u*ao_overlap_abs_grid(j,i).lt.thrsh_cycle_tc)cycle
int_fit = overlap_gauss_r12_ao_with1s(B_center, beta, r, expo_fit, i, j)
tmp += coef * coef_fit * int_fit
enddo
@ -325,7 +317,7 @@ BEGIN_PROVIDER [ double precision, v_ij_u_cst_mu_j1b_ng_1_test, (ao_num, ao_num,
!$OMP expo_gauss_j_mu_x, coef_gauss_j_mu_x, &
!$OMP List_comb_thr_b2_coef, List_comb_thr_b2_expo,List_comb_thr_b2_size, &
!$OMP List_comb_thr_b2_cent, v_ij_u_cst_mu_j1b_ng_1_test,ao_abs_comb_b2_j1b, &
!$OMP ao_overlap_abs_grid,ao_prod_center,ao_prod_sigma,dsqpi_3_2)
!$OMP ao_overlap_abs_grid,ao_prod_center,ao_prod_sigma,dsqpi_3_2,thrsh_cycle_tc)
!$OMP DO
do ipoint = 1, n_points_final_grid
r(1) = final_grid_points(1,ipoint)
@ -334,7 +326,7 @@ BEGIN_PROVIDER [ double precision, v_ij_u_cst_mu_j1b_ng_1_test, (ao_num, ao_num,
do i = 1, ao_num
do j = i, ao_num
if(dabs(ao_overlap_abs_grid(j,i)).lt.1.d-20)cycle
if(dabs(ao_overlap_abs_grid(j,i)).lt.thrsh_cycle_tc)cycle
tmp = 0.d0
@ -343,7 +335,7 @@ BEGIN_PROVIDER [ double precision, v_ij_u_cst_mu_j1b_ng_1_test, (ao_num, ao_num,
! --- --- ---
int_j1b = ao_abs_comb_b2_j1b(1,j,i)
if(dabs(int_j1b).lt.1.d-10) cycle
! if(dabs(int_j1b).lt.thrsh_cycle_tc) cycle
expo_fit = expo_good_j_mu_1gauss
int_fit = overlap_gauss_r12_ao(r, expo_fit, i, j)
tmp += int_fit
@ -356,7 +348,7 @@ BEGIN_PROVIDER [ double precision, v_ij_u_cst_mu_j1b_ng_1_test, (ao_num, ao_num,
coef = List_comb_thr_b2_coef (i_1s,j,i)
beta = List_comb_thr_b2_expo (i_1s,j,i)
int_j1b = ao_abs_comb_b2_j1b(i_1s,j,i)
if(dabs(coef)*dabs(int_j1b).lt.1.d-10)cycle
! if(dabs(coef)*dabs(int_j1b).lt.thrsh_cycle_tc)cycle
B_center(1) = List_comb_thr_b2_cent(1,i_1s,j,i)
B_center(2) = List_comb_thr_b2_cent(2,i_1s,j,i)
B_center(3) = List_comb_thr_b2_cent(3,i_1s,j,i)
@ -364,9 +356,9 @@ BEGIN_PROVIDER [ double precision, v_ij_u_cst_mu_j1b_ng_1_test, (ao_num, ao_num,
expo_fit = expo_good_j_mu_1gauss
coef_fit = 1.d0
coeftot = coef * coef_fit
if(dabs(coeftot).lt.1.d-15)cycle
if(dabs(coeftot).lt.thrsh_cycle_tc)cycle
call gaussian_product(beta,B_center,expo_fit,r,factor_ij_1s_u,beta_ij_u,center_ij_1s_u)
if(factor_ij_1s_u*ao_overlap_abs_grid(j,i).lt.1.d-15)cycle
if(factor_ij_1s_u*ao_overlap_abs_grid(j,i).lt.thrsh_cycle_tc)cycle
int_fit = overlap_gauss_r12_ao_with1s(B_center, beta, r, expo_fit, i, j)
tmp += coef * coef_fit * int_fit
! enddo

42
src/ao_many_one_e_ints/listj1b_sorted.irp.f
View File

@ -3,15 +3,16 @@
&BEGIN_PROVIDER [ integer, max_List_comb_thr_b2_size]
implicit none
integer :: i_1s,i,j,ipoint
double precision :: coef,beta,center(3),int_j1b,thr
double precision :: coef,beta,center(3),int_j1b
double precision :: r(3),weight,dist
thr = 1.d-15
List_comb_thr_b2_size = 0
print*,'List_all_comb_b2_size = ',List_all_comb_b2_size
! pause
do i = 1, ao_num
do j = i, ao_num
do i_1s = 1, List_all_comb_b2_size
coef = List_all_comb_b2_coef (i_1s)
if(dabs(coef).lt.1.d-15)cycle
if(dabs(coef).lt.thrsh_cycle_tc)cycle
beta = List_all_comb_b2_expo (i_1s)
beta = max(beta,1.d-12)
center(1:3) = List_all_comb_b2_cent(1:3,i_1s)
@ -24,7 +25,7 @@
dist += ( center(3) - r(3) )*( center(3) - r(3) )
int_j1b += dabs(aos_in_r_array_extra_transp(ipoint,i) * aos_in_r_array_extra_transp(ipoint,j))*dexp(-beta*dist) * weight
enddo
if(dabs(coef)*dabs(int_j1b).gt.thr)then
if(dabs(coef)*dabs(int_j1b).gt.thrsh_cycle_tc)then
List_comb_thr_b2_size(j,i) += 1
endif
enddo
@ -40,6 +41,7 @@
list(i) = maxval(List_comb_thr_b2_size(:,i))
enddo
max_List_comb_thr_b2_size = maxval(list)
print*,'max_List_comb_thr_b2_size = ',max_List_comb_thr_b2_size
END_PROVIDER
@ -49,16 +51,15 @@ END_PROVIDER
&BEGIN_PROVIDER [ double precision, ao_abs_comb_b2_j1b, ( max_List_comb_thr_b2_size ,ao_num, ao_num)]
implicit none
integer :: i_1s,i,j,ipoint,icount
double precision :: coef,beta,center(3),int_j1b,thr
double precision :: coef,beta,center(3),int_j1b
double precision :: r(3),weight,dist
thr = 1.d-15
ao_abs_comb_b2_j1b = 10000000.d0
do i = 1, ao_num
do j = i, ao_num
icount = 0
do i_1s = 1, List_all_comb_b2_size
coef = List_all_comb_b2_coef (i_1s)
if(dabs(coef).lt.1.d-12)cycle
if(dabs(coef).lt.thrsh_cycle_tc)cycle
beta = List_all_comb_b2_expo (i_1s)
center(1:3) = List_all_comb_b2_cent(1:3,i_1s)
int_j1b = 0.d0
@ -70,7 +71,7 @@ END_PROVIDER
dist += ( center(3) - r(3) )*( center(3) - r(3) )
int_j1b += dabs(aos_in_r_array_extra_transp(ipoint,i) * aos_in_r_array_extra_transp(ipoint,j))*dexp(-beta*dist) * weight
enddo
if(dabs(coef)*dabs(int_j1b).gt.thr)then
if(dabs(coef)*dabs(int_j1b).gt.thrsh_cycle_tc)then
icount += 1
List_comb_thr_b2_coef(icount,j,i) = coef
List_comb_thr_b2_expo(icount,j,i) = beta
@ -98,17 +99,17 @@ END_PROVIDER
&BEGIN_PROVIDER [ integer, max_List_comb_thr_b3_size]
implicit none
integer :: i_1s,i,j,ipoint
double precision :: coef,beta,center(3),int_j1b,thr
double precision :: coef,beta,center(3),int_j1b
double precision :: r(3),weight,dist
thr = 1.d-15
List_comb_thr_b3_size = 0
print*,'List_all_comb_b3_size = ',List_all_comb_b3_size
do i = 1, ao_num
do j = 1, ao_num
do i_1s = 1, List_all_comb_b3_size
coef = List_all_comb_b3_coef (i_1s)
beta = List_all_comb_b3_expo (i_1s)
center(1:3) = List_all_comb_b3_cent(1:3,i_1s)
if(dabs(coef).lt.thr)cycle
if(dabs(coef).lt.thrsh_cycle_tc)cycle
int_j1b = 0.d0
do ipoint = 1, n_points_extra_final_grid
r(1:3) = final_grid_points_extra(1:3,ipoint)
@ -118,7 +119,7 @@ END_PROVIDER
dist += ( center(3) - r(3) )*( center(3) - r(3) )
int_j1b += dabs(aos_in_r_array_extra_transp(ipoint,i) * aos_in_r_array_extra_transp(ipoint,j))*dexp(-beta*dist) * weight
enddo
if(dabs(coef)*dabs(int_j1b).gt.thr)then
if(dabs(coef)*dabs(int_j1b).gt.thrsh_cycle_tc)then
List_comb_thr_b3_size(j,i) += 1
endif
enddo
@ -144,9 +145,8 @@ END_PROVIDER
&BEGIN_PROVIDER [ double precision, ao_abs_comb_b3_j1b, ( max_List_comb_thr_b3_size ,ao_num, ao_num)]
implicit none
integer :: i_1s,i,j,ipoint,icount
double precision :: coef,beta,center(3),int_j1b,thr
double precision :: coef,beta,center(3),int_j1b
double precision :: r(3),weight,dist
thr = 1.d-15
ao_abs_comb_b3_j1b = 10000000.d0
do i = 1, ao_num
do j = 1, ao_num
@ -156,7 +156,7 @@ END_PROVIDER
beta = List_all_comb_b3_expo (i_1s)
beta = max(beta,1.d-12)
center(1:3) = List_all_comb_b3_cent(1:3,i_1s)
if(dabs(coef).lt.thr)cycle
if(dabs(coef).lt.thrsh_cycle_tc)cycle
int_j1b = 0.d0
do ipoint = 1, n_points_extra_final_grid
r(1:3) = final_grid_points_extra(1:3,ipoint)
@ -166,7 +166,7 @@ END_PROVIDER
dist += ( center(3) - r(3) )*( center(3) - r(3) )
int_j1b += dabs(aos_in_r_array_extra_transp(ipoint,i) * aos_in_r_array_extra_transp(ipoint,j))*dexp(-beta*dist) * weight
enddo
if(dabs(coef)*dabs(int_j1b).gt.thr)then
if(dabs(coef)*dabs(int_j1b).gt.thrsh_cycle_tc)then
icount += 1
List_comb_thr_b3_coef(icount,j,i) = coef
List_comb_thr_b3_expo(icount,j,i) = beta
@ -177,15 +177,5 @@ END_PROVIDER
enddo
enddo
! do i = 1, ao_num
! do j = 1, i-1
! do icount = 1, List_comb_thr_b3_size(j,i)
! List_comb_thr_b3_coef(icount,j,i) = List_comb_thr_b3_coef(icount,i,j)
! List_comb_thr_b3_expo(icount,j,i) = List_comb_thr_b3_expo(icount,i,j)
! List_comb_thr_b3_cent(1:3,icount,j,i) = List_comb_thr_b3_cent(1:3,icount,i,j)
! enddo
! enddo
! enddo
END_PROVIDER

12
src/bi_ort_ints/three_body_ijmk.irp.f
View File

@ -27,7 +27,7 @@ BEGIN_PROVIDER [ double precision, three_e_4_idx_direct_bi_ort, (mo_num, mo_num,
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,k,m,integral) &
!$OMP SHARED (mo_num,three_e_4_idx_direct_bi_ort)
!$OMP DO SCHEDULE (dynamic)
!$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
do i = 1, mo_num
do k = 1, mo_num
do j = 1, mo_num
@ -74,7 +74,7 @@ BEGIN_PROVIDER [ double precision, three_e_4_idx_cycle_1_bi_ort, (mo_num, mo_num
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,k,m,integral) &
!$OMP SHARED (mo_num,three_e_4_idx_cycle_1_bi_ort)
!$OMP DO SCHEDULE (dynamic)
!$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
do i = 1, mo_num
do k = 1, mo_num
do j = 1, mo_num
@ -121,7 +121,7 @@ BEGIN_PROVIDER [ double precision, three_e_4_idx_cycle_2_bi_ort, (mo_num, mo_num
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,k,m,integral) &
!$OMP SHARED (mo_num,three_e_4_idx_cycle_2_bi_ort)
!$OMP DO SCHEDULE (dynamic)
!$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
do i = 1, mo_num
do k = 1, mo_num
do j = 1, mo_num
@ -168,7 +168,7 @@ BEGIN_PROVIDER [ double precision, three_e_4_idx_exch23_bi_ort, (mo_num, mo_num,
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,k,m,integral) &
!$OMP SHARED (mo_num,three_e_4_idx_exch23_bi_ort)
!$OMP DO SCHEDULE (dynamic)
!$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
do i = 1, mo_num
do k = 1, mo_num
do j = 1, mo_num
@ -214,7 +214,7 @@ BEGIN_PROVIDER [ double precision, three_e_4_idx_exch13_bi_ort, (mo_num, mo_num,
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,k,m,integral) &
!$OMP SHARED (mo_num,three_e_4_idx_exch13_bi_ort)
!$OMP DO SCHEDULE (dynamic)
!$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
do i = 1, mo_num
do k = 1, mo_num
do j = 1, mo_num
@ -261,7 +261,7 @@ BEGIN_PROVIDER [ double precision, three_e_4_idx_exch12_bi_ort, (mo_num, mo_num,
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,k,m,integral) &
!$OMP SHARED (mo_num,three_e_4_idx_exch12_bi_ort)
!$OMP DO SCHEDULE (dynamic)
!$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
do i = 1, mo_num
do k = 1, mo_num
do j = 1, mo_num

12
src/bi_ort_ints/three_body_ijmkl.irp.f
View File

@ -26,7 +26,7 @@ BEGIN_PROVIDER [ double precision, three_e_5_idx_direct_bi_ort, (mo_num, mo_num,
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,k,m,l,integral) &
!$OMP SHARED (mo_num,three_e_5_idx_direct_bi_ort)
!$OMP DO SCHEDULE (dynamic)
!$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
do i = 1, mo_num
do k = 1, mo_num
do j = 1, mo_num
@ -75,7 +75,7 @@ BEGIN_PROVIDER [ double precision, three_e_5_idx_cycle_1_bi_ort, (mo_num, mo_num
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,k,m,l,integral) &
!$OMP SHARED (mo_num,three_e_5_idx_cycle_1_bi_ort)
!$OMP DO SCHEDULE (dynamic)
!$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
do i = 1, mo_num
do k = 1, mo_num
do j = 1, mo_num
@ -124,7 +124,7 @@ BEGIN_PROVIDER [ double precision, three_e_5_idx_cycle_2_bi_ort, (mo_num, mo_num
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,k,m,l,integral) &
!$OMP SHARED (mo_num,three_e_5_idx_cycle_2_bi_ort)
!$OMP DO SCHEDULE (dynamic)
!$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
do i = 1, mo_num
do k = 1, mo_num
do j = 1, mo_num
@ -173,7 +173,7 @@ BEGIN_PROVIDER [ double precision, three_e_5_idx_exch23_bi_ort, (mo_num, mo_num,
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,k,m,l,integral) &
!$OMP SHARED (mo_num,three_e_5_idx_exch23_bi_ort)
!$OMP DO SCHEDULE (dynamic)
!$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
do i = 1, mo_num
do k = 1, mo_num
do j = 1, mo_num
@ -222,7 +222,7 @@ BEGIN_PROVIDER [ double precision, three_e_5_idx_exch13_bi_ort, (mo_num, mo_num,
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,k,m,l,integral) &
!$OMP SHARED (mo_num,three_e_5_idx_exch13_bi_ort)
!$OMP DO SCHEDULE (dynamic)
!$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
do i = 1, mo_num
do k = 1, mo_num
do j = 1, mo_num
@ -271,7 +271,7 @@ BEGIN_PROVIDER [ double precision, three_e_5_idx_exch12_bi_ort, (mo_num, mo_num,
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i,j,k,m,l,integral) &
!$OMP SHARED (mo_num,three_e_5_idx_exch12_bi_ort)
!$OMP DO SCHEDULE (dynamic)
!$OMP DO SCHEDULE (dynamic) COLLAPSE(2)
do i = 1, mo_num
do k = 1, mo_num
do j = 1, mo_num

1
src/bi_ort_ints/total_twoe_pot.irp.f
View File

@ -57,6 +57,7 @@ BEGIN_PROVIDER [double precision, ao_two_e_tc_tot, (ao_num, ao_num, ao_num, ao_n
PROVIDE ao_tc_sym_two_e_pot_in_map
!!! TODO :: OPENMP
do j = 1, ao_num
do l = 1, ao_num
do i = 1, ao_num

92
src/bi_ortho_mos/overlap.irp.f
View File

@ -12,32 +12,27 @@
double precision :: accu_d, accu_nd
double precision, allocatable :: tmp(:,:)
! TODO : re do the DEGEMM
! overlap_bi_ortho = 0.d0
! do i = 1, mo_num
! do k = 1, mo_num
! do m = 1, ao_num
! do n = 1, ao_num
! overlap_bi_ortho(k,i) += ao_overlap(n,m) * mo_l_coef(n,k) * mo_r_coef(m,i)
! enddo
! enddo
! enddo
! enddo
overlap_bi_ortho = 0.d0
do i = 1, mo_num
do k = 1, mo_num
do m = 1, ao_num
do n = 1, ao_num
overlap_bi_ortho(k,i) += ao_overlap(n,m) * mo_l_coef(n,k) * mo_r_coef(m,i)
enddo
enddo
enddo
enddo
! allocate( tmp(mo_num,ao_num) )
!
! ! tmp <-- L.T x S_ao
! call dgemm( "T", "N", mo_num, ao_num, ao_num, 1.d0 &
! , mo_l_coef, size(mo_l_coef, 1), ao_overlap, size(ao_overlap, 1) &
! , 0.d0, tmp, size(tmp, 1) )
!
! ! S <-- tmp x R
! call dgemm( "N", "N", mo_num, mo_num, ao_num, 1.d0 &
! , tmp, size(tmp, 1), mo_r_coef, size(mo_r_coef, 1) &
! , 0.d0, overlap_bi_ortho, size(overlap_bi_ortho, 1) )
!
! deallocate( tmp )
allocate( tmp(mo_num,ao_num) )
! tmp <-- L.T x S_ao
call dgemm( "T", "N", mo_num, ao_num, ao_num, 1.d0 &
, mo_l_coef(1,1), size(mo_l_coef, 1), ao_overlap(1,1), size(ao_overlap, 1) &
, 0.d0, tmp(1,1), size(tmp, 1) )
! S <-- tmp x R
call dgemm( "N", "N", mo_num, mo_num, ao_num, 1.d0 &
, tmp(1,1), size(tmp, 1), mo_r_coef(1,1), size(mo_r_coef, 1) &
, 0.d0, overlap_bi_ortho(1,1), size(overlap_bi_ortho, 1) )
deallocate(tmp)
do i = 1, mo_num
overlap_diag_bi_ortho(i) = overlap_bi_ortho(i,i)
@ -84,20 +79,41 @@ END_PROVIDER
END_DOC
implicit none
integer :: i, j, p, q
integer :: i, j, p, q
double precision, allocatable :: tmp(:,:)
overlap_mo_r = 0.d0
overlap_mo_l = 0.d0
do i = 1, mo_num
do j = 1, mo_num
do p = 1, ao_num
do q = 1, ao_num
overlap_mo_r(j,i) += mo_r_coef(q,i) * mo_r_coef(p,j) * ao_overlap(q,p)
overlap_mo_l(j,i) += mo_l_coef(q,i) * mo_l_coef(p,j) * ao_overlap(q,p)
enddo
enddo
enddo
enddo
!overlap_mo_r = 0.d0
!overlap_mo_l = 0.d0
!do i = 1, mo_num
! do j = 1, mo_num
! do p = 1, ao_num
! do q = 1, ao_num
! overlap_mo_r(j,i) += mo_r_coef(q,i) * mo_r_coef(p,j) * ao_overlap(q,p)
! overlap_mo_l(j,i) += mo_l_coef(q,i) * mo_l_coef(p,j) * ao_overlap(q,p)
! enddo
! enddo
! enddo
!enddo
allocate( tmp(mo_num,ao_num) )
tmp = 0.d0
call dgemm( "T", "N", mo_num, ao_num, ao_num, 1.d0 &
, mo_r_coef(1,1), size(mo_r_coef, 1), ao_overlap(1,1), size(ao_overlap, 1) &
, 0.d0, tmp(1,1), size(tmp, 1) )
call dgemm( "N", "N", mo_num, mo_num, ao_num, 1.d0 &
, tmp(1,1), size(tmp, 1), mo_r_coef(1,1), size(mo_r_coef, 1) &
, 0.d0, overlap_mo_r(1,1), size(overlap_mo_r, 1) )
tmp = 0.d0
call dgemm( "T", "N", mo_num, ao_num, ao_num, 1.d0 &
, mo_l_coef(1,1), size(mo_l_coef, 1), ao_overlap(1,1), size(ao_overlap, 1) &
, 0.d0, tmp(1,1), size(tmp, 1) )
call dgemm( "N", "N", mo_num, mo_num, ao_num, 1.d0 &
, tmp(1,1), size(tmp, 1), mo_l_coef(1,1), size(mo_l_coef, 1) &
, 0.d0, overlap_mo_l(1,1), size(overlap_mo_l, 1) )
deallocate(tmp)
END_PROVIDER

55
src/davidson/EZFIO.cfg
View File

@ -1,71 +1,18 @@
[threshold_davidson]
type: Threshold
doc: Thresholds of Davidson's algorithm if threshold_davidson_from_pt2 is false.
interface: ezfio,provider,ocaml
default: 1.e-10
[threshold_nonsym_davidson]
type: Threshold
doc: Thresholds of non-symetric Davidson's algorithm
interface: ezfio,provider,ocaml
default: 1.e-10
[threshold_davidson_from_pt2]
type: logical
doc: Thresholds of Davidson's algorithm is set to E(rPT2)*threshold_davidson_from_pt2
interface: ezfio,provider,ocaml
default: false
[n_states_diag]
type: States_number
doc: Controls the number of states to consider during the Davdison diagonalization. The number of states is n_states * n_states_diag
default: 4
interface: ezfio,ocaml
[davidson_sze_max]
type: Strictly_positive_int
doc: Number of micro-iterations before re-contracting
default: 15
interface: ezfio,provider,ocaml
[state_following]
type: logical
doc: If |true|, the states are re-ordered to match the input states
default: False
interface: ezfio,provider,ocaml
[disk_based_davidson]
type: logical
doc: If |true|, a memory-mapped file may be used to store the W and S2 vectors if not enough RAM is available
default: True
interface: ezfio,provider,ocaml
[csf_based]
type: logical
doc: If |true|, use the CSF-based algorithm
default: False
interface: ezfio,provider,ocaml
[distributed_davidson]
type: logical
doc: If |true|, use the distributed algorithm
default: True
interface: ezfio,provider,ocaml
[only_expected_s2]
type: logical
doc: If |true|, use filter out all vectors with bad |S^2| values
default: True
interface: ezfio,provider,ocaml
[n_det_max_full]
type: Det_number_max
doc: Maximum number of determinants where |H| is fully diagonalized
interface: ezfio,provider,ocaml
default: 1000
[without_diagonal]
type: logical
doc: If |true|, don't use denominator
default: False
interface: ezfio,provider,ocaml

1
src/davidson/NEED
View File

@ -1 +1,2 @@
csf
davidson_keywords

15
src/davidson/davidson_parallel.irp.f
View File

@ -546,21 +546,6 @@ end
BEGIN_PROVIDER [ integer, nthreads_davidson ]
implicit none
BEGIN_DOC
! Number of threads for Davidson
END_DOC
nthreads_davidson = nproc
character*(32) :: env
call getenv('QP_NTHREADS_DAVIDSON',env)
if (trim(env) /= '') then
read(env,*) nthreads_davidson
call write_int(6,nthreads_davidson,'Target number of threads for <Psi|H|Psi>')
endif
END_PROVIDER
integer function zmq_put_N_states_diag(zmq_to_qp_run_socket,worker_id)
use f77_zmq
implicit none

23
src/davidson/diagonalization_hs2_dressed.irp.f
View File

@ -14,15 +14,6 @@ BEGIN_PROVIDER [ character*(64), diag_algorithm ]
endif
END_PROVIDER
BEGIN_PROVIDER [ double precision, threshold_davidson_pt2 ]
implicit none
BEGIN_DOC
! Threshold of Davidson's algorithm, using PT2 as a guide
END_DOC
threshold_davidson_pt2 = threshold_davidson
END_PROVIDER
BEGIN_PROVIDER [ integer, dressed_column_idx, (N_states) ]
@ -66,7 +57,7 @@ subroutine davidson_diag_hs2(dets_in,u_in,s2_out,dim_in,energies,sze,N_st,N_st_d
double precision, allocatable :: H_jj(:)
double precision, external :: diag_H_mat_elem, diag_S_mat_elem
integer :: i,k
integer :: i,k,l
ASSERT (N_st > 0)
ASSERT (sze > 0)
ASSERT (Nint > 0)
@ -87,9 +78,14 @@ subroutine davidson_diag_hs2(dets_in,u_in,s2_out,dim_in,energies,sze,N_st,N_st_d
if (dressing_state > 0) then
do k=1,N_st
do i=1,sze
H_jj(i) += u_in(i,k) * dressing_column_h(i,k)
H_jj(i) += u_in(i,k) * dressing_column_h(i,k)
enddo
!l = dressed_column_idx(k)
!H_jj(l) += u_in(l,k) * dressing_column_h(l,k)
enddo
endif
@ -465,8 +461,9 @@ subroutine davidson_diag_hjj_sjj(dets_in,u_in,H_jj,s2_out,energies,dim_in,sze,N_
integer :: lwork, info
double precision, allocatable :: work(:)
! y = h
y = h_p
y = h
!y = h_p
lwork = -1
allocate(work(1))
call dsygv(1,'V','U',shift2,y,size(y,1), &

541
src/davidson/diagonalization_nonsym_h_dressed.irp.f Normal file
View File

@ -0,0 +1,541 @@
! ---
subroutine davidson_diag_nonsym_h(dets_in, u_in, dim_in, energies, sze, N_st, N_st_diag, Nint, dressing_state, converged)
BEGIN_DOC
!
! non-sym Davidson diagonalization.
!
! dets_in : bitmasks corresponding to determinants
!
! u_in : guess coefficients on the various states. Overwritten on exit
!
! dim_in : leftmost dimension of u_in
!
! sze : Number of determinants
!
! N_st : Number of eigenstates
!
! Initial guess vectors are not necessarily orthonormal
!
END_DOC
use bitmasks
implicit none
integer, intent(in) :: dim_in, sze, N_st, N_st_diag, Nint
integer, intent(in) :: dressing_state
integer(bit_kind), intent(in) :: dets_in(Nint,2,sze)
logical, intent(out) :: converged
double precision, intent(out) :: energies(N_st_diag)
double precision, intent(inout) :: u_in(dim_in,N_st_diag)
integer :: i, k, l
double precision :: f
double precision, allocatable :: H_jj(:)
double precision, external :: diag_H_mat_elem
ASSERT (N_st > 0)
ASSERT (sze > 0)
ASSERT (Nint > 0)
ASSERT (Nint == N_int)
PROVIDE mo_two_e_integrals_in_map
allocate(H_jj(sze))
H_jj(1) = diag_H_mat_elem(dets_in(1,1,1), Nint)
!$OMP PARALLEL DEFAULT(NONE) &
!$OMP SHARED(sze, H_jj, dets_in, Nint) &
!$OMP PRIVATE(i)
!$OMP DO SCHEDULE(static)
do i = 2, sze
H_jj(i) = diag_H_mat_elem(dets_in(1,1,i), Nint)
enddo
!$OMP END DO
!$OMP END PARALLEL
if(dressing_state > 0) then
do k = 1, N_st
do l = 1, N_st
f = overlap_states_inv(k,l)
!do i = 1, N_det
! H_jj(i) += f * dressing_delta(i,k) * psi_coef(i,l)
do i = 1, dim_in
H_jj(i) += f * dressing_delta(i,k) * u_in(i,l)
enddo
enddo
enddo
endif
call davidson_diag_nonsym_hjj(dets_in, u_in, H_jj, energies, dim_in, sze, N_st, N_st_diag, Nint, dressing_state, converged)
deallocate(H_jj)
end subroutine davidson_diag_nonsym_h
! ---
subroutine davidson_diag_nonsym_hjj(dets_in, u_in, H_jj, energies, dim_in, sze, N_st, N_st_diag_in, Nint, dressing_state, converged)
BEGIN_DOC
!
! non-sym Davidson diagonalization with specific diagonal elements of the H matrix
!
! H_jj : specific diagonal H matrix elements to diagonalize de Davidson
!
! dets_in : bitmasks corresponding to determinants
!
! u_in : guess coefficients on the various states. Overwritten on exit
!
! dim_in : leftmost dimension of u_in
!
! sze : Number of determinants
!
! N_st : Number of eigenstates
!
! N_st_diag_in : Number of states in which H is diagonalized. Assumed > sze
!
! Initial guess vectors are not necessarily orthonormal
!
END_DOC
include 'constants.include.F'
use bitmasks
use mmap_module
implicit none
integer, intent(in) :: dim_in, sze, N_st, N_st_diag_in, Nint
integer, intent(in) :: dressing_state
integer(bit_kind), intent(in) :: dets_in(Nint,2,sze)
double precision, intent(in) :: H_jj(sze)
double precision, intent(out) :: energies(N_st_diag_in)
logical, intent(inout) :: converged
double precision, intent(inout) :: u_in(dim_in,N_st_diag_in)
logical :: disk_based
character*(16384) :: write_buffer
integer :: i, j, k, l, m
integer :: iter, N_st_diag, itertot, shift, shift2, itermax, istate
integer :: nproc_target
integer :: order(N_st_diag_in)
integer :: maxab
double precision :: rss
double precision :: cmax
double precision :: to_print(2,N_st)
double precision :: r1, r2
double precision :: f
double precision, allocatable :: y(:,:), h(:,:), lambda(:)
double precision, allocatable :: s_tmp(:,:), u_tmp(:,:)
double precision, allocatable :: residual_norm(:)
double precision, allocatable :: U(:,:), overlap(:,:)
double precision, pointer :: W(:,:)
double precision, external :: u_dot_u
N_st_diag = N_st_diag_in
!DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: U, W, y, h, lambda
if(N_st_diag*3 > sze) then
print *, 'error in Davidson :'
print *, 'Increase n_det_max_full to ', N_st_diag*3
stop -1
endif
itermax = max(2, min(davidson_sze_max, sze/N_st_diag)) + 1
itertot = 0
if(state_following) then
allocate(overlap(N_st_diag*itermax, N_st_diag*itermax))
else
allocate(overlap(1,1)) ! avoid 'if' for deallocate
endif
overlap = 0.d0
PROVIDE nuclear_repulsion expected_s2 psi_bilinear_matrix_order psi_bilinear_matrix_order_reverse threshold_davidson_pt2 threshold_davidson_from_pt2
PROVIDE threshold_nonsym_davidson
call write_time(6)
write(6,'(A)') ''
write(6,'(A)') 'Davidson Diagonalization'
write(6,'(A)') '------------------------'
write(6,'(A)') ''
! Find max number of cores to fit in memory
! -----------------------------------------
nproc_target = nproc
maxab = max(N_det_alpha_unique, N_det_beta_unique) + 1
m=1
disk_based = .False.
call resident_memory(rss)
do
r1 = 8.d0 * &! bytes
( dble(sze)*(N_st_diag*itermax) &! U
+ 1.0d0*dble(sze*m)*(N_st_diag*itermax) &! W
+ 3.0d0*(N_st_diag*itermax)**2 &! h,y,s_tmp
+ 1.d0*(N_st_diag*itermax) &! lambda
+ 1.d0*(N_st_diag) &! residual_norm
! In H_u_0_nstates_zmq
+ 2.d0*(N_st_diag*N_det) &! u_t, v_t, on collector
+ 2.d0*(N_st_diag*N_det) &! u_t, v_t, on slave
+ 0.5d0*maxab &! idx0 in H_u_0_nstates_openmp_work_*
+ nproc_target * &! In OMP section
( 1.d0*(N_int*maxab) &! buffer
+ 3.5d0*(maxab) ) &! singles_a, singles_b, doubles, idx
) / 1024.d0**3
if(nproc_target == 0) then
call check_mem(r1, irp_here)
nproc_target = 1
exit
endif
if(r1+rss < qp_max_mem) then
exit
endif
if(itermax > 4) then
itermax = itermax - 1
else if(m==1 .and. disk_based_davidson) then
m = 0
disk_based = .True.
itermax = 6
else
nproc_target = nproc_target - 1
endif
enddo
nthreads_davidson = nproc_target
TOUCH nthreads_davidson
call write_int(6, N_st, 'Number of states')
call write_int(6, N_st_diag, 'Number of states in diagonalization')
call write_int(6, sze, 'Number of determinants')
call write_int(6, nproc_target, 'Number of threads for diagonalization')
call write_double(6, r1, 'Memory(Gb)')
if(disk_based) then
print *, 'Using swap space to reduce RAM'
endif
!---------------
write(6,'(A)') ''
write_buffer = '====='
do i = 1, N_st
write_buffer = trim(write_buffer)//' ================ ==========='
enddo
write(6, '(A)') write_buffer(1:6+41*N_st)
write_buffer = 'Iter'
do i = 1, N_st
write_buffer = trim(write_buffer)//' Energy Residual '
enddo
write(6,'(A)') write_buffer(1:6+41*N_st)
write_buffer = '====='
do i = 1, N_st
write_buffer = trim(write_buffer)//' ================ ==========='
enddo
write(6,'(A)') write_buffer(1:6+41*N_st)
if(disk_based) then
! Create memory-mapped files for W and S
type(c_ptr) :: ptr_w, ptr_s
integer :: fd_s, fd_w
call mmap(trim(ezfio_work_dir)//'davidson_w', (/int(sze,8),int(N_st_diag*itermax,8)/),&
8, fd_w, .False., ptr_w)
call c_f_pointer(ptr_w, w, (/sze,N_st_diag*itermax/))
else
allocate(W(sze,N_st_diag*itermax))
endif
allocate( &
! Large
U(sze,N_st_diag*itermax), &
! Small
h(N_st_diag*itermax,N_st_diag*itermax), &
y(N_st_diag*itermax,N_st_diag*itermax), &
s_tmp(N_st_diag*itermax,N_st_diag*itermax), &
residual_norm(N_st_diag), &
lambda(N_st_diag*itermax), &
u_tmp(N_st,N_st_diag))
h = 0.d0
U = 0.d0
y = 0.d0
s_tmp = 0.d0
ASSERT (N_st > 0)
ASSERT (N_st_diag >= N_st)
ASSERT (sze > 0)
ASSERT (Nint > 0)
ASSERT (Nint == N_int)
! Davidson iterations
! ===================
converged = .False.
do k = N_st+1, N_st_diag
do i = 1, sze
call random_number(r1)
call random_number(r2)
r1 = dsqrt(-2.d0*dlog(r1))
r2 = dtwo_pi*r2
u_in(i,k) = r1*dcos(r2) * u_in(i,k-N_st)
enddo
u_in(k,k) = u_in(k,k) + 10.d0
enddo
do k = 1, N_st_diag
call normalize(u_in(1,k), sze)
enddo
do k = 1, N_st_diag
do i = 1, sze
U(i,k) = u_in(i,k)
enddo
enddo
do while (.not.converged)
itertot = itertot + 1
if(itertot == 8) then
exit
endif
do iter = 1, itermax-1
shift = N_st_diag*(iter-1)
shift2 = N_st_diag*iter
! if( (iter > 1) .or. (itertot == 1) ) then
! Gram-Schmidt to orthogonalize all new guess with the previous vectors
call ortho_qr(U, size(U, 1), sze, shift2)
call ortho_qr(U, size(U, 1), sze, shift2)
! Compute |W_k> = \sum_i |i><i|H|u_k>
! -----------------------------------
if( (sze > 100000) .and. distributed_davidson ) then
call H_u_0_nstates_zmq (W(1,shift+1), U(1,shift+1), N_st_diag, sze)
else
call H_u_0_nstates_openmp(W(1,shift+1), U(1,shift+1), N_st_diag, sze)
endif
! else
! ! Already computed in update below
! continue
! endif
if(dressing_state > 0) then
call dgemm( 'T', 'N', N_st, N_st_diag, sze, 1.d0 &
, psi_coef, size(psi_coef, 1), U(1, shift+1), size(U, 1) &
, 0.d0, u_tmp, size(u_tmp, 1))
do istate = 1, N_st_diag
do k = 1, N_st
do l = 1, N_st
f = overlap_states_inv(k,l)
do i = 1, sze
W(i,shift+istate) += f * dressing_delta(i,k) * u_tmp(l,istate)
enddo
enddo
enddo
enddo
endif
! Compute h_kl = <u_k | W_l> = <u_k| H |u_l>
! -------------------------------------------
call dgemm( 'T', 'N', shift2, shift2, sze, 1.d0 &
, U, size(U, 1), W, size(W, 1) &
, 0.d0, h, size(h, 1))
! Diagonalize h
! ---------------
call diag_nonsym_right(shift2, h(1,1), size(h, 1), y(1,1), size(y, 1), lambda(1), size(lambda, 1))
if (state_following) then
overlap = -1.d0
do k = 1, shift2
do i = 1, shift2
overlap(k,i) = dabs(y(k,i))
enddo
enddo
do k = 1, N_st
cmax = -1.d0
do i = 1, N_st
if(overlap(i,k) > cmax) then
cmax = overlap(i,k)
order(k) = i
endif
enddo
do i = 1, N_st_diag
overlap(order(k),i) = -1.d0
enddo
enddo
overlap = y
do k = 1, N_st
l = order(k)
if (k /= l) then
y(1:shift2,k) = overlap(1:shift2,l)
endif
enddo
do k = 1, N_st
overlap(k,1) = lambda(k)
enddo
endif
! Express eigenvectors of h in the determinant basis
! --------------------------------------------------
call dgemm( 'N', 'N', sze, N_st_diag, shift2, 1.d0 &
, U, size(U, 1), y, size(y, 1) &
, 0.d0, U(1,shift2+1), size(U, 1))
do k = 1, N_st_diag
call normalize(U(1,shift2+k), sze)
enddo
call dgemm( 'N', 'N', sze, N_st_diag, shift2, 1.d0 &
, W, size(W, 1), y, size(y, 1) &
, 0.d0, W(1,shift2+1), size(W,1))
! Compute residual vector and davidson step
! -----------------------------------------
!$OMP PARALLEL DO DEFAULT(SHARED) PRIVATE(i,k)
do k = 1, N_st_diag
do i = 1, sze
U(i,shift2+k) = (lambda(k) * U(i,shift2+k) - W(i,shift2+k)) / max(H_jj(i)-lambda(k), 1.d-2)
enddo
if(k <= N_st) then
residual_norm(k) = u_dot_u(U(1,shift2+k), sze)
to_print(1,k) = lambda(k) + nuclear_repulsion
to_print(2,k) = residual_norm(k)
endif
enddo
!$OMP END PARALLEL DO
if((itertot>1).and.(iter == 1)) then
!don't print
continue
else
write(*, '(1X, I3, 1X, 100(1X, F16.10, 1X, E11.3))') iter-1, to_print(1:2,1:N_st)
endif
! Check convergence
if(iter > 1) then
if(threshold_davidson_from_pt2) then
converged = dabs(maxval(residual_norm(1:N_st))) < threshold_davidson_pt2
else
converged = dabs(maxval(residual_norm(1:N_st))) < threshold_nonsym_davidson
endif
endif
do k = 1, N_st
if(residual_norm(k) > 1.d8) then
print *, 'Davidson failed'
stop -1
endif
enddo
if(converged) then
exit
endif
logical, external :: qp_stop
if(qp_stop()) then
converged = .True.
exit
endif
enddo
! Re-contract U and update W
! --------------------------------
call dgemm( 'N', 'N', sze, N_st_diag, shift2, 1.d0 &
, W, size(W, 1), y, size(y, 1) &
, 0.d0, u_in, size(u_in, 1))
do k = 1, N_st_diag
do i = 1, sze
W(i,k) = u_in(i,k)
enddo
enddo
call dgemm( 'N', 'N', sze, N_st_diag, shift2, 1.d0 &
, U, size(U, 1), y, size(y, 1), 0.d0 &
, u_in, size(u_in, 1))
do k = 1, N_st_diag
do i = 1, sze
U(i,k) = u_in(i,k)
enddo
enddo
enddo
call nullify_small_elements(sze, N_st_diag, U, size(U, 1), threshold_davidson_pt2)
do k = 1, N_st_diag
do i = 1, sze
u_in(i,k) = U(i,k)
enddo
enddo
do k = 1, N_st_diag
energies(k) = lambda(k)
enddo
write_buffer = '======'
do i = 1, N_st
write_buffer = trim(write_buffer)//' ================ ==========='
enddo
write(6,'(A)') trim(write_buffer)
write(6,'(A)') ''
call write_time(6)
if(disk_based) then
! Remove temp files
integer, external :: getUnitAndOpen
call munmap( (/int(sze,8),int(N_st_diag*itermax,8)/), 8, fd_w, ptr_w )
fd_w = getUnitAndOpen(trim(ezfio_work_dir)//'davidson_w','r')
close(fd_w,status='delete')
else
deallocate(W)
endif
deallocate ( &
residual_norm, &
U, overlap, &
h, y, s_tmp, &
lambda, &
u_tmp &
)
FREE nthreads_davidson
end subroutine davidson_diag_nonsym_hjj
! ---

40
src/davidson/overlap_states.irp.f Normal file
View File

@ -0,0 +1,40 @@
! ---
BEGIN_PROVIDER [ double precision, overlap_states, (N_states,N_states) ]
&BEGIN_PROVIDER [ double precision, overlap_states_inv, (N_states,N_states) ]
BEGIN_DOC
!
! S_kl = ck.T x cl
! = psi_coef(:,k).T x psi_coef(:,l)
!
END_DOC
implicit none
integer :: i
double precision :: o_tmp
if(N_states == 1) then
o_tmp = 0.d0
do i = 1, N_det
o_tmp = o_tmp + psi_coef(i,1) * psi_coef(i,1)
enddo
overlap_states (1,1) = o_tmp
overlap_states_inv(1,1) = 1.d0 / o_tmp
else
call dgemm( 'T', 'N', N_states, N_states, N_det, 1.d0 &
, psi_coef, size(psi_coef, 1), psi_coef, size(psi_coef, 1) &
, 0.d0, overlap_states, size(overlap_states, 1) )
call get_inverse(overlap_states, N_states, N_states, overlap_states_inv, N_states)
endif
END_PROVIDER
! ---

188
src/davidson_dressed/nonsym_diagonalize_ci.irp.f Normal file
View File

@ -0,0 +1,188 @@
! ---
BEGIN_PROVIDER [ double precision, CI_energy_nonsym_dressed, (N_states_diag) ]
BEGIN_DOC
! N_states lowest eigenvalues of the CI matrix
END_DOC
implicit none
integer :: j
character*(8) :: st
call write_time(6)
do j = 1, min(N_det, N_states_diag)
CI_energy_nonsym_dressed(j) = CI_electronic_energy_nonsym_dressed(j) + nuclear_repulsion
enddo
do j = 1, min(N_det, N_states)
write(st, '(I4)') j
call write_double(6, CI_energy_nonsym_dressed(j), 'Energy of state '//trim(st))
enddo
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, CI_electronic_energy_nonsym_dressed, (N_states_diag) ]
&BEGIN_PROVIDER [ double precision, CI_eigenvectors_nonsym_dressed, (N_det,N_states_diag) ]
BEGIN_DOC
! Eigenvectors/values of the CI matrix
END_DOC
implicit none
logical :: converged
integer :: i, j, k
integer :: i_other_state
integer :: i_state
logical, allocatable :: good_state_array(:)
integer, allocatable :: index_good_state_array(:)
double precision, allocatable :: eigenvectors(:,:), eigenvalues(:)
PROVIDE threshold_nonsym_davidson nthreads_davidson
! Guess values for the "N_states" states of the CI_eigenvectors_nonsym_dressed
do j = 1, min(N_states, N_det)
do i = 1, N_det
CI_eigenvectors_nonsym_dressed(i,j) = psi_coef(i,j)
enddo
enddo
do j = min(N_states, N_det)+1, N_states_diag
do i = 1, N_det
CI_eigenvectors_nonsym_dressed(i,j) = 0.d0
enddo
enddo
! ---
if(diag_algorithm == "Davidson") then
ASSERT(n_states_diag .lt. n_states)
do j = 1, min(N_states, N_det)
do i = 1, N_det
CI_eigenvectors_nonsym_dressed(i,j) = psi_coef(i,j)
enddo
enddo
converged = .False.
call davidson_diag_nonsym_h( psi_det, CI_eigenvectors_nonsym_dressed &
, size(CI_eigenvectors_nonsym_dressed, 1) &
, CI_electronic_energy_nonsym_dressed &
, N_det, min(N_det, N_states), min(N_det, N_states_diag), N_int, 1, converged )
else if(diag_algorithm == "Lapack") then
allocate(eigenvectors(size(H_matrix_nonsym_dressed, 1),N_det))
allocate(eigenvalues(N_det))
call diag_nonsym_right( N_det, H_matrix_nonsym_dressed, size(H_matrix_nonsym_dressed, 1) &
, eigenvectors, size(eigenvectors, 1), eigenvalues, size(eigenvalues, 1) )
CI_electronic_energy_nonsym_dressed(:) = 0.d0
! Select the "N_states_diag" states of lowest energy
do j = 1, min(N_det, N_states_diag)
do i = 1, N_det
CI_eigenvectors_nonsym_dressed(i,j) = eigenvectors(i,j)
enddo
CI_electronic_energy_nonsym_dressed(j) = eigenvalues(j)
enddo
deallocate(eigenvectors, eigenvalues)
! --- ---
endif
! ---
END_PROVIDER
! ---
subroutine diagonalize_CI_nonsym_dressed()
BEGIN_DOC
! Replace the coefficients of the CI states by the coefficients of the
! eigenstates of the CI matrix
END_DOC
implicit none
integer :: i, j
PROVIDE dressing_delta
do j = 1, N_states
do i = 1, N_det
psi_coef(i,j) = CI_eigenvectors_nonsym_dressed(i,j)
enddo
enddo
SOFT_TOUCH psi_coef
end subroutine diagonalize_CI_nonsym_dressed
! ---
BEGIN_PROVIDER [ double precision, H_matrix_nonsym_dressed, (N_det,N_det) ]
BEGIN_DOC
! Dressed H with Delta_ij
END_DOC
implicit none
integer :: i, j, l, k
double precision :: f
H_matrix_nonsym_dressed(1:N_det,1:N_det) = h_matrix_all_dets(1:N_det,1:N_det)
if(N_states == 1) then
! !symmetric formula
! l = dressed_column_idx(1)
! f = 1.0d0/psi_coef(l,1)
! do i=1,N_det
! h_matrix_nonsym_dressed(i,l) += dressing_column_h(i,1) *f
! h_matrix_nonsym_dressed(l,i) += dressing_column_h(i,1) *f
! enddo
! l = dressed_column_idx(1)
! f = 1.0d0 / psi_coef(l,1)
! do j = 1, N_det
! H_matrix_nonsym_dressed(j,l) += f * dressing_delta(j,1)
! enddo
k = 1
l = 1
f = overlap_states_inv(k,l)
do j = 1, N_det
do i = 1, N_det
H_matrix_nonsym_dressed(i,j) = H_matrix_nonsym_dressed(i,j) + f * dressing_delta(i,k) * psi_coef(j,l)
enddo
enddo
else
do k = 1, N_states
do l = 1, N_states
f = overlap_states_inv(k,l)
do j = 1, N_det
do i = 1, N_det
H_matrix_nonsym_dressed(i,j) = H_matrix_nonsym_dressed(i,j) + f * dressing_delta(i,k) * psi_coef(j,l)
enddo
enddo
enddo
enddo
endif
END_PROVIDER
! ---

54
src/davidson_keywords/EZFIO.cfg Normal file
View File

@ -0,0 +1,54 @@
[threshold_davidson]
type: Threshold
doc: Thresholds of Davidson's algorithm if threshold_davidson_from_pt2 is false.
interface: ezfio,provider,ocaml
default: 1.e-10
[threshold_nonsym_davidson]
type: Threshold
doc: Thresholds of non-symetric Davidson's algorithm
interface: ezfio,provider,ocaml
default: 1.e-10
[davidson_sze_max]
type: Strictly_positive_int
doc: Number of micro-iterations before re-contracting
default: 15
interface: ezfio,provider,ocaml
[state_following]
type: logical
doc: If |true|, the states are re-ordered to match the input states
default: False
interface: ezfio,provider,ocaml
[disk_based_davidson]
type: logical
doc: If |true|, a memory-mapped file may be used to store the W and S2 vectors if not enough RAM is availabl
default: True
interface: ezfio,provider,ocaml
[n_states_diag]
type: States_number
doc: Controls the number of states to consider during the Davdison diagonalization. The number of states is n_states * n_states_diag
default: 4
interface: ezfio,ocaml
[n_det_max_full]
type: Det_number_max
doc: Maximum number of determinants where |H| is fully diagonalized
interface: ezfio,provider,ocaml
default: 1000
[threshold_davidson_from_pt2]
type: logical
doc: Thresholds of Davidson's algorithm is set to E(rPT2)*threshold_davidson_from_pt2
interface: ezfio,provider,ocaml
default: false
[distributed_davidson]
type: logical
doc: If |true|, use the distributed algorithm
default: True
interface: ezfio,provider,ocaml

1
src/davidson_keywords/NEED Normal file
View File

@ -0,0 +1 @@
ezfio_files

5
src/davidson_keywords/README.rst Normal file
View File

@ -0,0 +1,5 @@
=================
davidson_keywords
=================
Keywords used for Davidson algorithms.

10
src/davidson/input.irp.f → src/davidson_keywords/input.irp.f
View File

@ -1,3 +1,6 @@
! ---
BEGIN_PROVIDER [ integer, n_states_diag ]
implicit none
BEGIN_DOC
@ -8,11 +11,11 @@ BEGIN_PROVIDER [ integer, n_states_diag ]
PROVIDE ezfio_filename
if (mpi_master) then
call ezfio_has_davidson_n_states_diag(has)
call ezfio_has_davidson_keywords_n_states_diag(has)
if (has) then
call ezfio_get_davidson_n_states_diag(n_states_diag)
call ezfio_get_davidson_keywords_n_states_diag(n_states_diag)
else
print *, 'davidson/n_states_diag not found in EZFIO file'
print *, 'davidson_keywords/n_states_diag not found in EZFIO file'
stop 1
endif
n_states_diag = max(2,N_states * N_states_diag)
@ -32,3 +35,4 @@ BEGIN_PROVIDER [ integer, n_states_diag ]
END_PROVIDER
! ---

33
src/davidson_keywords/usef.irp.f Normal file
View File

@ -0,0 +1,33 @@
use bitmasks
use f77_zmq
! ---
BEGIN_PROVIDER [ integer, nthreads_davidson ]
implicit none
BEGIN_DOC
! Number of threads for Davidson
END_DOC
nthreads_davidson = nproc
character*(32) :: env
call getenv('QP_NTHREADS_DAVIDSON',env)
if (trim(env) /= '') then
read(env,*) nthreads_davidson
call write_int(6,nthreads_davidson,'Target number of threads for <Psi|H|Psi>')
endif
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, threshold_davidson_pt2 ]
implicit none
BEGIN_DOC
! Threshold of Davidson's algorithm, using PT2 as a guide
END_DOC
threshold_davidson_pt2 = threshold_davidson
END_PROVIDER
! ---

2
src/davidson_undressed/null_dressing_vector.irp.f
View File

@ -1,10 +1,12 @@
BEGIN_PROVIDER [ double precision, dressing_column_h, (N_det,N_states) ]
&BEGIN_PROVIDER [ double precision, dressing_column_s, (N_det,N_states) ]
&BEGIN_PROVIDER [ double precision, dressing_delta , (N_det,N_states) ]
implicit none
BEGIN_DOC
! Null dressing vectors
END_DOC
dressing_column_h(:,:) = 0.d0
dressing_column_s(:,:) = 0.d0
dressing_delta (:,:) = 0.d0
END_PROVIDER

5
src/determinants/spindeterminants.ezfio_config
View File

@ -9,8 +9,11 @@ spindeterminants
psi_det_beta integer*8 (spindeterminants_n_int*spindeterminants_bit_kind/8,spindeterminants_n_det_beta)
psi_coef_matrix_rows integer (spindeterminants_n_det)
psi_coef_matrix_columns integer (spindeterminants_n_det)
psi_coef_matrix_values double precision (spindeterminants_n_det,spindeterminants_n_states)
psi_coef_matrix_values double precision (spindeterminants_n_det,spindeterminants_n_states)
psi_left_coef_matrix_values double precision (spindeterminants_n_det,spindeterminants_n_states)
n_svd_coefs integer
n_svd_alpha integer
n_svd_beta integer
psi_svd_alpha double precision (spindeterminants_n_det_alpha,spindeterminants_n_svd_coefs,spindeterminants_n_states)
psi_svd_beta double precision (spindeterminants_n_det_beta,spindeterminants_n_svd_coefs,spindeterminants_n_states)
psi_svd_coefs double precision (spindeterminants_n_svd_coefs,spindeterminants_n_states)

30
src/kohn_sham/print_mos.irp.f Normal file
View File

@ -0,0 +1,30 @@
program print_mos
implicit none
integer :: i,nx
double precision :: r(3), xmax, dx, accu
double precision, allocatable :: mos_array(:)
double precision:: alpha,envelop
allocate(mos_array(mo_num))
xmax = 5.d0
nx = 1000
dx=xmax/dble(nx)
r = 0.d0
alpha = 0.5d0
do i = 1, nx
call give_all_mos_at_r(r,mos_array)
accu = mos_array(3)**2+mos_array(4)**2+mos_array(5)**2
accu = dsqrt(accu)
envelop = (1.d0 - dexp(-alpha * r(3)**2))
write(33,'(100(F16.10,X))')r(3), mos_array(1), mos_array(2), accu, envelop
r(3) += dx
enddo
end
double precision function f_mu(x)
implicit none
double precision, intent(in) :: x
end

70
src/non_h_ints_mu/grad_squared.irp.f
View File

@ -2,7 +2,7 @@
! ---
! TODO : strong optmization : write the loops in a different way
! : for each couple of AO, the gaussian product are done once for all
! : for each couple of AO, the gaussian product are done once for all
BEGIN_PROVIDER [ double precision, gradu_squared_u_ij_mu, (ao_num, ao_num, n_points_final_grid) ]
@ -20,14 +20,14 @@ BEGIN_PROVIDER [ double precision, gradu_squared_u_ij_mu, (ao_num, ao_num, n_poi
! gradu_squared_u_ij_mu = -0.50 x \int r2 \phi_i(2) \phi_j(2) [ v1^2 v2^2 ((grad_1 u12)^2 + (grad_2 u12^2)]) + u12^2 v2^2 (grad_1 v1)^2 + 2 u12 v1 v2^2 (grad_1 u12) . (grad_1 v1) ]
! = -0.25 x v1^2 \int r2 \phi_i(2) \phi_j(2) [1 - erf(mu r12)]^2 v2^2
! + -0.50 x (grad_1 v1)^2 \int r2 \phi_i(2) \phi_j(2) u12^2 v2^2
! + -1.00 x v1 (grad_1 v1) \int r2 \phi_i(2) \phi_j(2) (grad_1 u12) v2^2
! + -1.00 x v1 (grad_1 v1) \int r2 \phi_i(2) \phi_j(2) (grad_1 u12) v2^2
! = v1^2 x int2_grad1u2_grad2u2_j1b2
! + -0.5 x (grad_1 v1)^2 x int2_u2_j1b2
! + -1.0 X V1 x (grad_1 v1) \cdot [ int2_u_grad1u_j1b2 x r - int2_u_grad1u_x_j1b ]
!
!
END_DOC
implicit none
integer :: ipoint, i, j, m, igauss
double precision :: x, y, z, r(3), delta, coef
@ -100,7 +100,7 @@ BEGIN_PROVIDER [ double precision, gradu_squared_u_ij_mu, (ao_num, ao_num, n_poi
call wall_time(time1)
print*, ' Wall time for gradu_squared_u_ij_mu = ', time1 - time0
END_PROVIDER
END_PROVIDER
! ---
@ -151,7 +151,7 @@ END_PROVIDER
!
! deallocate(ac_mat)
!
!END_PROVIDER
!END_PROVIDER
! ---
@ -214,12 +214,12 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao_loop, (ao_num, ao_num, ao_nu
call wall_time(time1)
print*, ' Wall time for tc_grad_square_ao_loop = ', time1 - time0
END_PROVIDER
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, grad12_j12, (ao_num, ao_num, n_points_final_grid) ]
implicit none
integer :: ipoint, i, j, m, igauss
double precision :: r(3), delta, coef
@ -267,7 +267,7 @@ BEGIN_PROVIDER [ double precision, grad12_j12, (ao_num, ao_num, n_points_final_g
call wall_time(time1)
print*, ' Wall time for grad12_j12 = ', time1 - time0
END_PROVIDER
END_PROVIDER
! ---
@ -297,12 +297,12 @@ BEGIN_PROVIDER [ double precision, u12sq_j1bsq, (ao_num, ao_num, n_points_final_
call wall_time(time1)
print*, ' Wall time for u12sq_j1bsq = ', time1 - time0
END_PROVIDER
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, u12_grad1_u12_j1b_grad1_j1b, (ao_num, ao_num, n_points_final_grid) ]
implicit none
integer :: ipoint, i, j, m, igauss
double precision :: x, y, z
@ -347,7 +347,7 @@ BEGIN_PROVIDER [ double precision, u12_grad1_u12_j1b_grad1_j1b, (ao_num, ao_num,
call wall_time(time1)
print*, ' Wall time for u12_grad1_u12_j1b_grad1_j1b = ', time1 - time0
END_PROVIDER
END_PROVIDER
! ---
@ -370,26 +370,18 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao, (ao_num, ao_num, ao_num, ao
if(read_tc_integ) then
open(unit=11, form="unformatted", file='tc_grad_square_ao', action="read")
do i = 1, ao_num
do j = 1, ao_num
do k = 1, ao_num
do l = 1, ao_num
read(11) tc_grad_square_ao(l,k,j,i)
enddo
enddo
enddo
enddo
open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/tc_grad_square_ao', action="read")
read(11) tc_grad_square_ao
close(11)
else
allocate(b_mat(n_points_final_grid,ao_num,ao_num), tmp(ao_num,ao_num,n_points_final_grid))
b_mat = 0.d0
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i, k, ipoint) &
!$OMP PRIVATE (i, k, ipoint) &
!$OMP SHARED (aos_in_r_array_transp, b_mat, ao_num, n_points_final_grid, final_weight_at_r_vector)
!$OMP DO SCHEDULE (static)
do i = 1, ao_num
@ -401,11 +393,11 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao, (ao_num, ao_num, ao_num, ao
enddo
!$OMP END DO
!$OMP END PARALLEL
tmp = 0.d0
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (j, l, ipoint) &
!$OMP PRIVATE (j, l, ipoint) &
!$OMP SHARED (tmp, ao_num, n_points_final_grid, u12sq_j1bsq, u12_grad1_u12_j1b_grad1_j1b, grad12_j12)
!$OMP DO SCHEDULE (static)
do ipoint = 1, n_points_final_grid
@ -417,25 +409,25 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao, (ao_num, ao_num, ao_num, ao
enddo
!$OMP END DO
!$OMP END PARALLEL
tc_grad_square_ao = 0.d0
call dgemm( "N", "N", ao_num*ao_num, ao_num*ao_num, n_points_final_grid, 1.d0 &
, tmp(1,1,1), ao_num*ao_num, b_mat(1,1,1), n_points_final_grid &
, 1.d0, tc_grad_square_ao, ao_num*ao_num)
deallocate(tmp, b_mat)
call sum_A_At(tc_grad_square_ao(1,1,1,1), ao_num*ao_num)
!!$OMP PARALLEL &
!!$OMP DEFAULT (NONE) &
!!$OMP PRIVATE (i, j, k, l) &
!!$OMP PRIVATE (i, j, k, l) &
!!$OMP SHARED (ac_mat, tc_grad_square_ao, ao_num)
!!$OMP DO SCHEDULE (static)
! do j = 1, ao_num
! do l = 1, ao_num
! do i = 1, ao_num
! do k = 1, ao_num
! tc_grad_square_ao(k,i,l,j) = ac_mat(k,i,l,j) + ac_mat(l,j,k,i)
! tc_grad_square_ao(k,i,l,j) = ac_mat(k,i,l,j) + ac_mat(l,j,k,i)
! enddo
! enddo
! enddo
@ -444,23 +436,17 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao, (ao_num, ao_num, ao_num, ao
!!$OMP END PARALLEL
endif
if(write_tc_integ) then
open(unit=11, form="unformatted", file='tc_grad_square_ao', action="write")
do i = 1, ao_num
do j = 1, ao_num
do k = 1, ao_num
do l = 1, ao_num
write(11) tc_grad_square_ao(l,k,j,i)
enddo
enddo
enddo
enddo
if(write_tc_integ.and.mpi_master) then
open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/tc_grad_square_ao', action="write")
call ezfio_set_work_empty(.False.)
write(11) tc_grad_square_ao
close(11)
call ezfio_set_tc_keywords_io_tc_integ('Read')
endif
call wall_time(time1)
print*, ' Wall time for tc_grad_square_ao = ', time1 - time0
END_PROVIDER
END_PROVIDER
! ---

76
src/non_h_ints_mu/grad_squared_manu.irp.f
View File

@ -17,29 +17,21 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao_test, (ao_num, ao_num, ao_nu
call wall_time(time0)
if(read_tc_integ) then
open(unit=11, form="unformatted", file='tc_grad_square_ao_test', action="read")
do i = 1, ao_num
do j = 1, ao_num
do k = 1, ao_num
do l = 1, ao_num
read(11) tc_grad_square_ao_test(l,k,j,i)
enddo
enddo
enddo
enddo
open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/tc_grad_square_ao_test', action="read")
read(11) tc_grad_square_ao_test
close(11)
else
provide u12sq_j1bsq_test u12_grad1_u12_j1b_grad1_j1b_test grad12_j12_test
allocate(b_mat(n_points_final_grid,ao_num,ao_num), tmp(ao_num,ao_num,n_points_final_grid))
b_mat = 0.d0
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i, k, ipoint) &
!$OMP PRIVATE (i, k, ipoint) &
!$OMP SHARED (aos_in_r_array_transp, b_mat, ao_num, n_points_final_grid, final_weight_at_r_vector)
!$OMP DO SCHEDULE (static)
do i = 1, ao_num
@ -51,11 +43,11 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao_test, (ao_num, ao_num, ao_nu
enddo
!$OMP END DO
!$OMP END PARALLEL
tmp = 0.d0
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (j, l, ipoint) &
!$OMP PRIVATE (j, l, ipoint) &
!$OMP SHARED (tmp, ao_num, n_points_final_grid, u12sq_j1bsq_test, u12_grad1_u12_j1b_grad1_j1b_test, grad12_j12_test)
!$OMP DO SCHEDULE (static)
do ipoint = 1, n_points_final_grid
@ -67,23 +59,23 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao_test, (ao_num, ao_num, ao_nu
enddo
!$OMP END DO
!$OMP END PARALLEL
tc_grad_square_ao_test = 0.d0
call dgemm( "N", "N", ao_num*ao_num, ao_num*ao_num, n_points_final_grid, 1.d0 &
, tmp(1,1,1), ao_num*ao_num, b_mat(1,1,1), n_points_final_grid &
, 1.d0, tc_grad_square_ao_test, ao_num*ao_num)
deallocate(tmp, b_mat)
call sum_A_At(tc_grad_square_ao_test(1,1,1,1), ao_num*ao_num)
!do i = 1, ao_num
! do j = 1, ao_num
! do k = i, ao_num
! do l = max(j,k), ao_num
! tc_grad_square_ao_test(i,j,k,l) = 0.5d0 * (tc_grad_square_ao_test(i,j,k,l) + tc_grad_square_ao_test(k,l,i,j))
! tc_grad_square_ao_test(k,l,i,j) = tc_grad_square_ao_test(i,j,k,l)
! end do
! !if (j.eq.k) then
! ! do l = j+1, ao_num
! ! tc_grad_square_ao_test(i,j,k,l) = 0.5d0 * (tc_grad_square_ao_test(i,j,k,l) + tc_grad_square_ao_test(k,l,i,j))
@ -95,14 +87,14 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao_test, (ao_num, ao_num, ao_nu
! ! tc_grad_square_ao_test(k,l,i,j) = tc_grad_square_ao_test(i,j,k,l)
! ! enddo
! !endif
! enddo
! enddo
!enddo
!tc_grad_square_ao_test = 2.d0 * tc_grad_square_ao_test
! !$OMP PARALLEL &
! !$OMP DEFAULT (NONE) &
! !$OMP PRIVATE (i, j, k, l) &
! !$OMP PRIVATE (i, j, k, l) &
! !$OMP SHARED (tc_grad_square_ao_test, ao_num)
! !$OMP DO SCHEDULE (static)
! integer :: ii
@ -121,10 +113,10 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao_test, (ao_num, ao_num, ao_nu
! print *, ' ii =', ii
! !$OMP END DO
! !$OMP END PARALLEL
! !$OMP PARALLEL &
! !$OMP DEFAULT (NONE) &
! !$OMP PRIVATE (i, j, k, l) &
! !$OMP PRIVATE (i, j, k, l) &
! !$OMP SHARED (tc_grad_square_ao_test, ao_num)
! !$OMP DO SCHEDULE (static)
! do j = 1, ao_num
@ -144,24 +136,18 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao_test, (ao_num, ao_num, ao_nu
endif
if(write_tc_integ) then
open(unit=11, form="unformatted", file='tc_grad_square_ao_test', action="write")
do i = 1, ao_num
do j = 1, ao_num
do k = 1, ao_num
do l = 1, ao_num
write(11) tc_grad_square_ao_test(l,k,j,i)
enddo
enddo
enddo
enddo
if(write_tc_integ.and.mpi_master) then
open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/tc_grad_square_ao_test', action="write")
call ezfio_set_work_empty(.False.)
write(11) tc_grad_square_ao_test
close(11)
call ezfio_set_tc_keywords_io_tc_integ('Read')
endif
call wall_time(time1)
print*, ' Wall time for tc_grad_square_ao_test = ', time1 - time0
END_PROVIDER
END_PROVIDER
! ---
@ -189,7 +175,7 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao_test_ref, (ao_num, ao_num, a
b_mat = 0.d0
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i, k, ipoint) &
!$OMP PRIVATE (i, k, ipoint) &
!$OMP SHARED (aos_in_r_array_transp, b_mat, ao_num, n_points_final_grid, final_weight_at_r_vector)
!$OMP DO SCHEDULE (static)
do i = 1, ao_num
@ -205,7 +191,7 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao_test_ref, (ao_num, ao_num, a
tmp = 0.d0
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (j, l, ipoint) &
!$OMP PRIVATE (j, l, ipoint) &
!$OMP SHARED (tmp, ao_num, n_points_final_grid, u12sq_j1bsq_test, u12_grad1_u12_j1b_grad1_j1b_test, grad12_j12_test)
!$OMP DO SCHEDULE (static)
do ipoint = 1, n_points_final_grid
@ -226,7 +212,7 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao_test_ref, (ao_num, ao_num, a
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i, j, k, l) &
!$OMP PRIVATE (i, j, k, l) &
!$OMP SHARED (ac_mat, tc_grad_square_ao_test_ref, ao_num)
!$OMP DO SCHEDULE (static)
do j = 1, ao_num
@ -246,7 +232,7 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao_test_ref, (ao_num, ao_num, a
call wall_time(time1)
print*, ' Wall time for tc_grad_square_ao_test_ref = ', time1 - time0
END_PROVIDER
END_PROVIDER
! ---
@ -276,12 +262,12 @@ BEGIN_PROVIDER [ double precision, u12sq_j1bsq_test, (ao_num, ao_num, n_points_f
call wall_time(time1)
print*, ' Wall time for u12sq_j1bsq_test = ', time1 - time0
END_PROVIDER
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, u12_grad1_u12_j1b_grad1_j1b_test, (ao_num, ao_num, n_points_final_grid) ]
implicit none
integer :: ipoint, i, j, m, igauss
double precision :: x, y, z
@ -328,12 +314,12 @@ BEGIN_PROVIDER [ double precision, u12_grad1_u12_j1b_grad1_j1b_test, (ao_num, ao
call wall_time(time1)
print*, ' Wall time for u12_grad1_u12_j1b_grad1_j1b_test = ', time1 - time0
END_PROVIDER
END_PROVIDER
! ---
BEGIN_PROVIDER [ double precision, grad12_j12_test, (ao_num, ao_num, n_points_final_grid) ]
implicit none
integer :: ipoint, i, j, m, igauss
double precision :: r(3), delta, coef
@ -381,7 +367,7 @@ BEGIN_PROVIDER [ double precision, grad12_j12_test, (ao_num, ao_num, n_points_fi
call wall_time(time1)
print*, ' Wall time for grad12_j12_test = ', time1 - time0
END_PROVIDER
END_PROVIDER
! ---

56
src/non_h_ints_mu/new_grad_tc.irp.f
View File

@ -36,16 +36,8 @@ BEGIN_PROVIDER [ double precision, int2_grad1_u12_ao, (ao_num, ao_num, n_points_
if(read_tc_integ) then
open(unit=11, form="unformatted", file='int2_grad1_u12_ao', action="read")
do m = 1, 3
do ipoint = 1, n_points_final_grid
do j = 1, ao_num
do i = 1, ao_num
read(11) int2_grad1_u12_ao(i,j,ipoint,m)
enddo
enddo
enddo
enddo
open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/int2_grad1_u12_ao', action="read")
read(11) int2_grad1_u12_ao
close(11)
else
@ -89,18 +81,12 @@ BEGIN_PROVIDER [ double precision, int2_grad1_u12_ao, (ao_num, ao_num, n_points_
endif
if(write_tc_integ) then
open(unit=11, form="unformatted", file='int2_grad1_u12_ao', action="write")
do m = 1, 3
do ipoint = 1, n_points_final_grid
do j = 1, ao_num
do i = 1, ao_num
write(11) int2_grad1_u12_ao(i,j,ipoint,m)
enddo
enddo
enddo
enddo
if(write_tc_integ.and.mpi_master) then
open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/int2_grad1_u12_ao', action="write")
call ezfio_set_work_empty(.False.)
write(11) int2_grad1_u12_ao
close(11)
call ezfio_set_tc_keywords_io_tc_integ('Read')
endif
call wall_time(time1)
@ -319,16 +305,8 @@ BEGIN_PROVIDER [double precision, tc_grad_and_lapl_ao, (ao_num, ao_num, ao_num,
if(read_tc_integ) then
open(unit=11, form="unformatted", file='tc_grad_and_lapl_ao', action="read")
do i = 1, ao_num
do j = 1, ao_num
do k = 1, ao_num
do l = 1, ao_num
read(11) tc_grad_and_lapl_ao(l,k,j,i)
enddo
enddo
enddo
enddo
open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/tc_grad_and_lapl_ao', action="read")
read(11) tc_grad_and_lapl_ao
close(11)
else
@ -388,18 +366,12 @@ BEGIN_PROVIDER [double precision, tc_grad_and_lapl_ao, (ao_num, ao_num, ao_num,
endif
if(write_tc_integ) then
open(unit=11, form="unformatted", file='tc_grad_and_lapl_ao', action="write")
do i = 1, ao_num
do j = 1, ao_num
do k = 1, ao_num
do l = 1, ao_num
write(11) tc_grad_and_lapl_ao(l,k,j,i)
enddo
enddo
enddo
enddo
if(write_tc_integ.and.mpi_master) then
open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/tc_grad_and_lapl_ao', action="write")
call ezfio_set_work_empty(.False.)
write(11) tc_grad_and_lapl_ao
close(11)
call ezfio_set_tc_keywords_io_tc_integ('Read')
endif
call wall_time(time1)

103
src/non_h_ints_mu/new_grad_tc_manu.irp.f
View File

@ -3,7 +3,7 @@ BEGIN_PROVIDER [ double precision, int2_grad1_u12_ao_test, (ao_num, ao_num, n_po
BEGIN_DOC
!
! int2_grad1_u12_ao_test(i,j,ipoint,:) = \int dr2 [-1 * \grad_r1 J(r1,r2)] \phi_i(r2) \phi_j(r2)
! int2_grad1_u12_ao_test(i,j,ipoint,:) = \int dr2 [-1 * \grad_r1 J(r1,r2)] \phi_i(r2) \phi_j(r2)
!
! where r1 = r(ipoint)
!
@ -15,9 +15,9 @@ BEGIN_PROVIDER [ double precision, int2_grad1_u12_ao_test, (ao_num, ao_num, n_po
! if J(r1,r2) = u12 x v1 x v2
!
! int2_grad1_u12_ao_test(i,j,ipoint,:) = v1 x [ 0.5 x \int dr2 [(r1 - r2) (erf(mu * r12)-1)r_12] v2 \phi_i(r2) \phi_j(r2) ]
! - \grad_1 v1 x [ \int dr2 u12 v2 \phi_i(r2) \phi_j(r2) ]
! = 0.5 v_1b(ipoint) * v_ij_erf_rk_cst_mu_j1b(i,j,ipoint) * r(:)
! - 0.5 v_1b(ipoint) * x_v_ij_erf_rk_cst_mu_j1b(i,j,ipoint,:)
! - \grad_1 v1 x [ \int dr2 u12 v2 \phi_i(r2) \phi_j(r2) ]
! = 0.5 v_1b(ipoint) * v_ij_erf_rk_cst_mu_j1b(i,j,ipoint) * r(:)
! - 0.5 v_1b(ipoint) * x_v_ij_erf_rk_cst_mu_j1b(i,j,ipoint,:)
! - v_1b_grad[:,ipoint] * v_ij_u_cst_mu_j1b(i,j,ipoint)
!
!
@ -35,25 +35,18 @@ BEGIN_PROVIDER [ double precision, int2_grad1_u12_ao_test, (ao_num, ao_num, n_po
if(read_tc_integ) then
open(unit=11, form="unformatted", file='int2_grad1_u12_ao_test', action="read")
do m = 1, 3
do ipoint = 1, n_points_final_grid
do j = 1, ao_num
do i = 1, ao_num
read(11) int2_grad1_u12_ao_test(i,j,ipoint,m)
enddo
enddo
enddo
enddo
open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/int2_grad1_u12_ao_test', action="read")
read(11) int2_grad1_u12_ao_test
close(11)
else
if(j1b_type .eq. 3) then
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)
z = final_grid_points(3,ipoint)
tmp0 = 0.5d0 * v_1b(ipoint)
tmp_x = v_1b_grad(1,ipoint)
tmp_y = v_1b_grad(2,ipoint)
@ -87,24 +80,18 @@ BEGIN_PROVIDER [ double precision, int2_grad1_u12_ao_test, (ao_num, ao_num, n_po
endif
if(write_tc_integ) then
open(unit=11, form="unformatted", file='int2_grad1_u12_ao_test', action="write")
do m = 1, 3
do ipoint = 1, n_points_final_grid
do j = 1, ao_num
do i = 1, ao_num
write(11) int2_grad1_u12_ao_test(i,j,ipoint,m)
enddo
enddo
enddo
enddo
if(write_tc_integ.and.mpi_master) then
open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/int2_grad1_u12_ao_test', action="write")
call ezfio_set_work_empty(.False.)
write(11) int2_grad1_u12_ao_test
close(11)
call ezfio_set_tc_keywords_io_tc_integ('Read')
endif
call wall_time(time1)
print*, ' Wall time for int2_grad1_u12_ao_test = ', time1 - time0
END_PROVIDER
END_PROVIDER
! ---
@ -114,9 +101,9 @@ BEGIN_PROVIDER [double precision, tc_grad_and_lapl_ao_test, (ao_num, ao_num, ao_
!
! tc_grad_and_lapl_ao_test(k,i,l,j) = < k l | -1/2 \Delta_1 u(r1,r2) - \grad_1 u(r1,r2) | ij >
!
! = 1/2 \int dr1 (phi_k(r1) \grad_r1 phi_i(r1) - phi_i(r1) \grad_r1 phi_k(r1)) . \int dr2 \grad_r1 u(r1,r2) \phi_l(r2) \phi_j(r2)
! = 1/2 \int dr1 (phi_k(r1) \grad_r1 phi_i(r1) - phi_i(r1) \grad_r1 phi_k(r1)) . \int dr2 \grad_r1 u(r1,r2) \phi_l(r2) \phi_j(r2)
!
! This is obtained by integration by parts.
! This is obtained by integration by parts.
!
END_DOC
@ -131,40 +118,32 @@ BEGIN_PROVIDER [double precision, tc_grad_and_lapl_ao_test, (ao_num, ao_num, ao_
call wall_time(time0)
if(read_tc_integ) then
open(unit=11, form="unformatted", file='tc_grad_and_lapl_ao_test', action="read")
do i = 1, ao_num
do j = 1, ao_num
do k = 1, ao_num
do l = 1, ao_num
read(11) tc_grad_and_lapl_ao_test(l,k,j,i)
enddo
enddo
enddo
enddo
open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/tc_grad_and_lapl_ao_test', action="read")
read(11) tc_grad_and_lapl_ao_test
close(11)
else
provide int2_grad1_u12_ao_test
provide int2_grad1_u12_ao_test
allocate(b_mat(n_points_final_grid,ao_num,ao_num,3), ac_mat(ao_num,ao_num,ao_num,ao_num))
b_mat = 0.d0
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i, k, ipoint, weight1, ao_i_r, ao_k_r) &
!$OMP SHARED (aos_in_r_array_transp, aos_grad_in_r_array_transp_bis, b_mat, &
!$OMP PRIVATE (i, k, ipoint, weight1, ao_i_r, ao_k_r) &
!$OMP SHARED (aos_in_r_array_transp, aos_grad_in_r_array_transp_bis, b_mat, &
!$OMP ao_num, n_points_final_grid, final_weight_at_r_vector)
!$OMP DO SCHEDULE (static)
do i = 1, ao_num
do k = 1, ao_num
do ipoint = 1, n_points_final_grid
weight1 = 0.5d0 * final_weight_at_r_vector(ipoint)
ao_i_r = aos_in_r_array_transp(ipoint,i)
ao_k_r = aos_in_r_array_transp(ipoint,k)
b_mat(ipoint,k,i,1) = weight1 * (ao_k_r * aos_grad_in_r_array_transp_bis(ipoint,i,1) - ao_i_r * aos_grad_in_r_array_transp_bis(ipoint,k,1))
b_mat(ipoint,k,i,2) = weight1 * (ao_k_r * aos_grad_in_r_array_transp_bis(ipoint,i,2) - ao_i_r * aos_grad_in_r_array_transp_bis(ipoint,k,2))
b_mat(ipoint,k,i,3) = weight1 * (ao_k_r * aos_grad_in_r_array_transp_bis(ipoint,i,3) - ao_i_r * aos_grad_in_r_array_transp_bis(ipoint,k,3))
@ -173,19 +152,19 @@ BEGIN_PROVIDER [double precision, tc_grad_and_lapl_ao_test, (ao_num, ao_num, ao_
enddo
!$OMP END DO
!$OMP END PARALLEL
ac_mat = 0.d0
do m = 1, 3
call dgemm( "N", "N", ao_num*ao_num, ao_num*ao_num, n_points_final_grid, 1.d0 &
, int2_grad1_u12_ao_test(1,1,1,m), ao_num*ao_num, b_mat(1,1,1,m), n_points_final_grid &
, 1.d0, ac_mat, ao_num*ao_num)
enddo
deallocate(b_mat)
!$OMP PARALLEL &
!$OMP DEFAULT (NONE) &
!$OMP PRIVATE (i, j, k, l) &
!$OMP PRIVATE (i, j, k, l) &
!$OMP SHARED (ac_mat, tc_grad_and_lapl_ao_test, ao_num)
!$OMP DO SCHEDULE (static)
do j = 1, ao_num
@ -199,29 +178,23 @@ BEGIN_PROVIDER [double precision, tc_grad_and_lapl_ao_test, (ao_num, ao_num, ao_
enddo
!$OMP END DO
!$OMP END PARALLEL
deallocate(ac_mat)
endif
if(write_tc_integ) then
open(unit=11, form="unformatted", file='tc_grad_and_lapl_ao_test', action="write")
do i = 1, ao_num
do j = 1, ao_num
do k = 1, ao_num
do l = 1, ao_num
write(11) tc_grad_and_lapl_ao_test(l,k,j,i)
enddo
enddo
enddo
enddo
if(write_tc_integ.and.mpi_master) then
open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/tc_grad_and_lapl_ao_test', action="write")
call ezfio_set_work_empty(.False.)
write(11) tc_grad_and_lapl_ao_test
close(11)
call ezfio_set_tc_keywords_io_tc_integ('Read')
endif
call wall_time(time1)
print*, ' Wall time for tc_grad_and_lapl_ao_test = ', time1 - time0
END_PROVIDER
END_PROVIDER
! ---

19
src/non_h_ints_mu/total_tc_int.irp.f
View File

@ -68,7 +68,26 @@ BEGIN_PROVIDER [double precision, ao_tc_int_chemist, (ao_num, ao_num, ao_num, ao
END_PROVIDER
BEGIN_PROVIDER [double precision, ao_tc_int_chemist_no_cycle, (ao_num, ao_num, ao_num, ao_num)]
! ---
implicit none
integer :: i, j, k, l
double precision :: wall1, wall0
print *, ' providing ao_tc_int_chemist_no_cycle ...'
call wall_time(wall0)
do j = 1, ao_num
do l = 1, ao_num
do i = 1, ao_num
do k = 1, ao_num
ao_tc_int_chemist_no_cycle(k,i,l,j) = tc_grad_square_ao(k,i,l,j) + tc_grad_and_lapl_ao(k,i,l,j) + ao_two_e_coul(k,i,l,j)
! ao_tc_int_chemist(k,i,l,j) = ao_two_e_coul(k,i,l,j)
enddo
enddo
enddo
enddo
call wall_time(wall1)
print *, ' wall time for ao_tc_int_chemist_no_cycle ', wall1 - wall0
END_PROVIDER
BEGIN_PROVIDER [double precision, ao_tc_int_chemist_test, (ao_num, ao_num, ao_num, ao_num)]

18
src/nuclei/write_pt_charges.py
View File

@ -21,7 +21,7 @@ def mv_in_ezfio(ezfio,tmp):
os.system(cmdmv)
# Getting the EZFIO
##Getting the EZFIO
EZFIO=sys.argv[1]
EZFIO=EZFIO.replace("/", "")
print(EZFIO)
@ -66,8 +66,20 @@ zip_in_ezfio(EZFIO,tmp)
tmp="pts_charge_coord"
fcoord = open(tmp,'w')
fcoord.write(" 2\n")
fcoord.write(" "+str(n_charges)+' 3\n')
#fcoord.write(" "+' 3 '+str(n_charges)+' \n')
if(n_charges < 10):
fcoord.write(" "+str(n_charges)+' 3\n')
elif(n_charges <100):
fcoord.write(" "+str(n_charges)+' 3\n')
elif(n_charges <1000):
fcoord.write(" "+str(n_charges)+' 3\n')
elif(n_charges <10000):
fcoord.write(" "+str(n_charges)+' 3\n')
elif(n_charges <100000):
fcoord.write(" "+str(n_charges)+' 3\n')
elif(n_charges <1000000):
fcoord.write(" "+str(n_charges)+' 3\n')
elif(n_charges <10000000):
fcoord.write(" "+str(n_charges)+' 3\n')
for i in range(n_charges):
fcoord.write(' '+coord_x[i]+'\n')
for i in range(n_charges):

557
src/tc_bi_ortho/dav_h_tc_s2.irp.f Normal file
View File

@ -0,0 +1,557 @@
! ---
subroutine davidson_hs2_nonsym_b1space(u_in, H_jj, s2_out,energies, sze, N_st, N_st_diag_in, n_it_max_dav, converged, hcalc)
use mmap_module
BEGIN_DOC
! Generic modified-Davidson diagonalization
!
! H_jj : specific diagonal H matrix elements to diagonalize de Davidson
!
! u_in : guess coefficients on the various states. Overwritten on exit by right eigenvectors
!
! sze : Number of determinants
!
! N_st : Number of eigenstates
!
! N_st_diag_in : Number of states in which H is diagonalized. Assumed > N_st
!
! Initial guess vectors are not necessarily orthonormal
!
! hcalc subroutine to compute W = H U (see routine hcalc_template for template of input/output)
!
! !!! WARNING !!! IT SEEMS THAT IF THE NUMBER OF MACRO ITERATIONS EXCEEDS n_it_max_dav,
!
! THE RECONTRACTION IS WRONG. YOU SHOULD CONSIDER CALLING MULTIPLE TIME THE ROUTINE
!
! SEE FOR INSTANCE IN tc_bi_ortho/tc_h_eigvectors.irp.f
END_DOC
implicit none
integer, intent(in) :: sze, N_st, N_st_diag_in, n_it_max_dav
double precision, intent(in) :: H_jj(sze)
logical, intent(inout) :: converged
double precision, intent(inout) :: u_in(sze,N_st_diag_in)
double precision, intent(out) :: energies(N_st)
double precision, intent(inout) :: s2_out(N_st)
external hcalc
character*(16384) :: write_buffer
integer :: iter, N_st_diag
integer :: i, j, k, l, m
integer :: iter2, itertot
logical :: disk_based
integer :: shift, shift2, itermax
integer :: nproc_target
integer :: order(N_st_diag_in)
double precision :: to_print(3,N_st)
double precision :: r1, r2, alpha
double precision :: cpu, wall
double precision :: cmax
double precision :: energy_shift(N_st_diag_in*davidson_sze_max)
double precision, allocatable :: U(:,:)
double precision, allocatable :: y(:,:), h(:,:), lambda(:), h_p(:,:), s2(:)
real, allocatable :: y_s(:,:)
double precision, allocatable :: s_(:,:), s_tmp(:,:)
double precision, allocatable :: residual_norm(:)
double precision :: lambda_tmp
integer, allocatable :: i_omax(:)
double precision, allocatable :: U_tmp(:), overlap(:), S_d(:,:)
double precision, allocatable :: W(:,:)
real, pointer :: S(:,:)
!double precision, pointer :: W(:,:)
double precision, external :: u_dot_v, u_dot_u
include 'constants.include.F'
N_st_diag = N_st_diag_in
! print*,'trial vector'
do i = 1, sze
if(isnan(u_in(i,1)))then
print*,'pb in input vector of davidson_general_ext_rout_nonsym_b1space'
print*,i,u_in(i,1)
stop
else if (dabs(u_in(i,1)).lt.1.d-16)then
u_in(i,1) = 0.d0
endif
enddo
!DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: U, W, S, y, y_s, S_d, h, lambda
if(N_st_diag*3 > sze) then
print *, 'error in Davidson :'
print *, 'Increase n_det_max_full to ', N_st_diag*3
stop -1
endif
itermax = max(2, min(davidson_sze_max, sze/N_st_diag)) + 1
provide threshold_nonsym_davidson
call write_time(6)
write(6,'(A)') ''
write(6,'(A)') 'Davidson Diagonalization'
write(6,'(A)') '------------------------'
write(6,'(A)') ''
! Find max number of cores to fit in memory
! -----------------------------------------
nproc_target = nproc
double precision :: rss
integer :: maxab
maxab = sze
m=1
disk_based = .False.
call resident_memory(rss)
do
r1 = 8.d0 * &! bytes
( dble(sze)*(N_st_diag*itermax) &! U
+ 1.5d0*dble(sze*m)*(N_st_diag*itermax) &! W, S
+ 4.5d0*(N_st_diag*itermax)**2 &! h,y,y_s,s_, s_tmp
+ 2.d0*(N_st_diag*itermax) &! s2,lambda
+ 1.d0*(N_st_diag) &! residual_norm
! In H_S2_u_0_nstates_zmq
+ 3.d0*(N_st_diag*N_det) &! u_t, v_t, s_t on collector
+ 3.d0*(N_st_diag*N_det) &! u_t, v_t, s_t on slave
+ 0.5d0*maxab &! idx0 in H_S2_u_0_nstates_openmp_work_*
+ nproc_target * &! In OMP section
( 1.d0*(N_int*maxab) &! buffer
+ 3.5d0*(maxab) ) &! singles_a, singles_b, doubles, idx
) / 1024.d0**3
if(nproc_target == 0) then
call check_mem(r1, irp_here)
nproc_target = 1
exit
endif
if(r1+rss < qp_max_mem) then
exit
endif
if(itermax > 4) then
itermax = itermax - 1
! else if (m==1.and.disk_based_davidson) then
! m = 0
! disk_based = .True.
! itermax = 6
else
nproc_target = nproc_target - 1
endif
enddo
nthreads_davidson = nproc_target
TOUCH nthreads_davidson
call write_int(6, N_st, 'Number of states')
call write_int(6, N_st_diag, 'Number of states in diagonalization')
call write_int(6, sze, 'Number of basis functions')
call write_int(6, nproc_target, 'Number of threads for diagonalization')
call write_double(6, r1, 'Memory(Gb)')
if(disk_based) then
print *, 'Using swap space to reduce RAM'
endif
!---------------
write(6,'(A)') ''
write_buffer = '====='
do i=1,N_st
write_buffer = trim(write_buffer)//' ================ =========== ==========='
enddo
write(6,'(A)') write_buffer(1:6+41*N_st)
write_buffer = 'Iter'
do i=1,N_st
write_buffer = trim(write_buffer)//' Energy S^2 Residual '
enddo
write(6,'(A)') write_buffer(1:6+41*N_st)
write_buffer = '====='
do i=1,N_st
write_buffer = trim(write_buffer)//' ================ =========== ==========='
enddo
write(6,'(A)') write_buffer(1:6+41*N_st)
! ---
allocate( W(sze,N_st_diag*itermax), S(sze,N_st_diag*itermax) )
allocate( &
! Large
U(sze,N_st_diag*itermax), &
S_d(sze,N_st_diag), &
! Small
h(N_st_diag*itermax,N_st_diag*itermax), &
h_p(N_st_diag*itermax,N_st_diag*itermax), &
y(N_st_diag*itermax,N_st_diag*itermax), &
s_(N_st_diag*itermax,N_st_diag*itermax), &
s_tmp(N_st_diag*itermax,N_st_diag*itermax), &
lambda(N_st_diag*itermax), &
residual_norm(N_st_diag), &
i_omax(N_st), &
s2(N_st_diag*itermax), &
y_s(N_st_diag*itermax,N_st_diag*itermax) &
)
U = 0.d0
h = 0.d0
y = 0.d0
s_ = 0.d0
s_tmp = 0.d0
lambda = 0.d0
residual_norm = 0.d0
ASSERT (N_st > 0)
ASSERT (N_st_diag >= N_st)
ASSERT (sze > 0)
! Davidson iterations
! ===================
converged = .False.
! Initialize from N_st to N_st_diag with gaussian random numbers
! to be sure to have overlap with any eigenvectors
do k = N_st+1, N_st_diag
u_in(k,k) = 10.d0
do i = 1, sze
call random_number(r1)
call random_number(r2)
r1 = dsqrt(-2.d0*dlog(r1))
r2 = dtwo_pi*r2
u_in(i,k) = r1*dcos(r2)
enddo
enddo
! Normalize all states
do k = 1, N_st_diag
call normalize(u_in(1,k), sze)
enddo
! Copy from the guess input "u_in" to the working vectors "U"
do k = 1, N_st_diag
do i = 1, sze
U(i,k) = u_in(i,k)
enddo
enddo
! ---
itertot = 0
! do while (.not.converged.or.itertot.le.n_it_max_dav)
integer :: iiii
do iiii = 1, n_it_max_dav
itertot = itertot + 1
if(itertot == 8) then
exit
endif
do iter = 1, itermax-1
shift = N_st_diag * (iter-1)
shift2 = N_st_diag * iter
if( (iter > 1) .or. (itertot == 1) ) then
! Gram-Schmidt to orthogonalize all new guess with the previous vectors
call ortho_qr(U, size(U, 1), sze, shift2)
call ortho_qr(U, size(U, 1), sze, shift2)
! W = H U
! call hcalc(W(1,shift+1), U(1,shift+1), N_st_diag, sze)
call hcalc(W(1,shift+1),S_d,U(1,shift+1),N_st_diag,sze)
S(1:sze,shift+1:shift+N_st_diag) = real(S_d(1:sze,1:N_st_diag))
else
! Already computed in update below
continue
endif
! Compute s_kl = <u_k | S_l> = <u_k| S2 |u_l>
! -------------------------------------------
!$OMP PARALLEL DO DEFAULT(SHARED) PRIVATE(i,j,k) COLLAPSE(2)
do j=1,shift2
do i=1,shift2
s_(i,j) = 0.d0
do k=1,sze
s_(i,j) = s_(i,j) + U(k,i) * dble(S(k,j))
enddo
enddo
enddo
!$OMP END PARALLEL DO
! Compute h_kl = <u_k | W_l> = <u_k| H |u_l>
! -------------------------------------------
call dgemm( 'T', 'N', shift2, shift2, sze, 1.d0 &
, U, size(U, 1), W, size(W, 1) &
, 0.d0, h, size(h, 1) )
! Penalty method
! --------------
if (s2_eig) then
h_p = s_
do k=1,shift2
h_p(k,k) = h_p(k,k) - expected_s2
enddo
if (only_expected_s2) then
alpha = 0.1d0
h_p = h + alpha*h_p
else
alpha = 0.0001d0
h_p = h + alpha*h_p
endif
else
h_p = h
alpha = 0.d0
endif
! Diagonalize h y = lambda y
! ---------------------------
call diag_nonsym_right(shift2, h_p(1,1), size(h_p, 1), y(1,1), size(y, 1), lambda(1), size(lambda, 1))
do k = 1, N_st_diag
! print*,'lambda(k) before = ',lambda(k)
lambda(k) = 0.d0
do l = 1, shift2
do m = 1, shift2
lambda(k) += y(m,k) * h(m,l) * y(l,k)
enddo
enddo
! print*,'lambda(k) new = ',lambda(k)
enddo
! Compute S2 for each eigenvector
! -------------------------------
call dgemm('N','N',shift2,shift2,shift2, &
1.d0, s_, size(s_,1), y, size(y,1), &
0.d0, s_tmp, size(s_tmp,1))
call dgemm('T','N',shift2,shift2,shift2, &
1.d0, y, size(y,1), s_tmp, size(s_tmp,1), &
0.d0, s_, size(s_,1))
do k=1,shift2
s2(k) = s_(k,k)
enddo
! Express eigenvectors of h in the determinant basis:
! ---------------------------------------------------
! y(:,k) = rk
! U(:,k) = Bk
! U(:,shift2+k) = Rk = Bk x rk
call dgemm( 'N', 'N', sze, N_st_diag, shift2, 1.d0 &
, U, size(U, 1), y, size(y, 1) &
, 0.d0, U(1,shift2+1), size(U, 1) )
do k = 1, N_st_diag
call normalize(U(1,shift2+k), sze)
enddo
! ---
! select the max overlap
!
! start test ------------------------------------------------------------------------
!
!double precision, allocatable :: Utest(:,:), Otest(:)
!allocate( Utest(sze,shift2), Otest(shift2) )
!call dgemm( 'N', 'N', sze, shift2, shift2, 1.d0 &
! , U, size(U, 1), y, size(y, 1), 0.d0, Utest(1,1), size(Utest, 1) )
!do k = 1, shift2
! call normalize(Utest(1,k), sze)
!enddo
!do j = 1, sze
! write(455, '(100(1X, F16.10))') (Utest(j,k), k=1,shift2)
!enddo
!do k = 1, shift2
! Otest(k) = 0.d0
! do i = 1, sze
! Otest(k) += Utest(i,k) * u_in(i,1)
! enddo
! Otest(k) = dabs(Otest(k))
! print *, ' Otest =', k, Otest(k), lambda(k)
!enddo
!deallocate(Utest, Otest)
!
! end test ------------------------------------------------------------------------
!
! TODO
! state_following is more efficient
do l = 1, N_st
allocate( overlap(N_st_diag) )
do k = 1, N_st_diag
overlap(k) = 0.d0
do i = 1, sze
overlap(k) = overlap(k) + U(i,shift2+k) * u_in(i,l)
enddo
overlap(k) = dabs(overlap(k))
!print *, ' overlap =', k, overlap(k)
enddo
lambda_tmp = 0.d0
do k = 1, N_st_diag
if(overlap(k) .gt. lambda_tmp) then
i_omax(l) = k
lambda_tmp = overlap(k)
endif
enddo
deallocate(overlap)
if(lambda_tmp .lt. 0.7d0) then
print *, ' very small overlap ...', l, i_omax(l)
print *, ' max overlap = ', lambda_tmp
stop
endif
if(i_omax(l) .ne. l) then
print *, ' !!! WARNONG !!!'
print *, ' index of state', l, i_omax(l)
endif
enddo
! y(:,k) = rk
! W(:,k) = H x Bk
! W(:,shift2+k) = H x Bk x rk
! = Wk
call dgemm( 'N', 'N', sze, N_st_diag, shift2, 1.d0 &
, W, size(W, 1), y, size(y, 1) &
, 0.d0, W(1,shift2+1), size(W, 1) )
! ---
! Compute residual vector and davidson step
! -----------------------------------------
!$OMP PARALLEL DO DEFAULT(SHARED) PRIVATE(i,k)
do k = 1, N_st_diag
do i = 1, sze
U(i,shift2+k) = (lambda(k) * U(i,shift2+k) - W(i,shift2+k)) / max(H_jj(i)-lambda(k), 1.d-2)
enddo
if(k <= N_st) then
l = k
residual_norm(k) = u_dot_u(U(1,shift2+l), sze)
to_print(1,k) = lambda(l)
to_print(2,k) = s2(l)
to_print(3,k) = residual_norm(l)
endif
enddo
!$OMP END PARALLEL DO
!residual_norm(1) = u_dot_u(U(1,shift2+1), sze)
!to_print(1,1) = lambda(1)
!to_print(2,1) = residual_norm(1)
if( (itertot > 1) .and. (iter == 1) ) then
!don't print
continue
else
write(*, '(1X, I3, 1X, 100(1X, F16.10, 1X, F16.10, 1X, F16.10))') iter-1, to_print(1:3,1:N_st)
endif
! Check convergence
if(iter > 1) then
converged = dabs(maxval(residual_norm(1:N_st))) < threshold_nonsym_davidson
endif
do k = 1, N_st
if(residual_norm(k) > 1.e8) then
print *, 'Davidson failed'
stop -1
endif
enddo
if(converged) then
exit
endif
logical, external :: qp_stop
if(qp_stop()) then
converged = .True.
exit
endif
enddo ! loop over iter
! Re-contract U and update W
! --------------------------------
call dgemm( 'N', 'N', sze, N_st_diag, shift2, 1.d0 &
, W, size(W, 1), y, size(y, 1) &
, 0.d0, u_in, size(u_in, 1) )
do k = 1, N_st_diag
do i = 1, sze
W(i,k) = u_in(i,k)
enddo
enddo
call dgemm( 'N', 'N', sze, N_st_diag, shift2, 1.d0 &
, U, size(U, 1), y, size(y, 1) &
, 0.d0, u_in, size(u_in, 1) )
do k = 1, N_st_diag
do i = 1, sze
U(i,k) = u_in(i,k)
enddo
enddo
call ortho_qr(U, size(U, 1), sze, N_st_diag)
call ortho_qr(U, size(U, 1), sze, N_st_diag)
do j = 1, N_st_diag
k = 1
do while( (k < sze) .and. (U(k,j) == 0.d0) )
k = k+1
enddo
if(U(k,j) * u_in(k,j) < 0.d0) then
do i = 1, sze
W(i,j) = -W(i,j)
enddo
endif
enddo
if(converged)exit
enddo ! loop over while
! ---
do k = 1, N_st
energies(k) = lambda(k)
s2_out(k) = s2(k)
enddo
write_buffer = '====='
do i = 1, N_st
write_buffer = trim(write_buffer)//' ================ ==========='
enddo
write(6,'(A)') trim(write_buffer)
write(6,'(A)') ''
call write_time(6)
deallocate(W)
deallocate(U, h, y, lambda, residual_norm, i_omax)
FREE nthreads_davidson
end subroutine davidson_general_ext_rout_nonsym_b1space
! ---

769
src/tc_bi_ortho/h_tc_s2_u0.irp.f Normal file
View File

@ -0,0 +1,769 @@
subroutine get_H_tc_s2_l0_r0(l_0,r_0,N_st,sze,energies, s2)
use bitmasks
implicit none
BEGIN_DOC
! Computes $e_0 = \langle l_0 | H | r_0\rangle$.
!
! Computes $s_0 = \langle l_0 | S^2 | r_0\rangle$.
!
! Assumes that the determinants are in psi_det
!
! istart, iend, ishift, istep are used in ZMQ parallelization.
END_DOC
integer, intent(in) :: N_st,sze
double precision, intent(in) :: l_0(sze,N_st), r_0(sze,N_st)
double precision, intent(out) :: energies(N_st), s2(N_st)
logical :: do_right
integer :: istate
double precision, allocatable :: s_0(:,:), v_0(:,:)
double precision :: u_dot_v, norm
allocate(s_0(sze,N_st), v_0(sze,N_st))
do_right = .True.
call H_tc_s2_u_0_opt(v_0,s_0,r_0,N_st,sze)
do istate = 1, N_st
norm = u_dot_v(l_0(1,istate),r_0(1,istate),sze)
energies(istate) = u_dot_v(l_0(1,istate),v_0(1,istate),sze)/norm
s2(istate) = u_dot_v(l_0(1,istate),s_0(1,istate),sze)/norm
enddo
end
subroutine H_tc_s2_u_0_opt(v_0,s_0,u_0,N_st,sze)
use bitmasks
implicit none
BEGIN_DOC
! Computes $v_0 = H | u_0\rangle$.
!
! Assumes that the determinants are in psi_det
!
! istart, iend, ishift, istep are used in ZMQ parallelization.
END_DOC
integer, intent(in) :: N_st,sze
double precision, intent(inout) :: v_0(sze,N_st), u_0(sze,N_st), s_0(sze,N_st)
logical :: do_right
do_right = .True.
call H_tc_s2_u_0_nstates_openmp(v_0,s_0,u_0,N_st,sze, do_right)
end
subroutine H_tc_s2_dagger_u_0_opt(v_0,s_0,u_0,N_st,sze)
use bitmasks
implicit none
BEGIN_DOC
! Computes $v_0 = H | u_0\rangle$.
!
! Assumes that the determinants are in psi_det
!
! istart, iend, ishift, istep are used in ZMQ parallelization.
END_DOC
integer, intent(in) :: N_st,sze
double precision, intent(inout) :: v_0(sze,N_st), u_0(sze,N_st), s_0(sze,N_st)
logical :: do_right
do_right = .False.
call H_tc_s2_u_0_nstates_openmp(v_0,s_0,u_0,N_st,sze, do_right)
end
subroutine H_tc_s2_u_0_nstates_openmp(v_0,s_0,u_0,N_st,sze, do_right)
use bitmasks
implicit none
BEGIN_DOC
! Computes $v_0 = H | u_0\rangle$.
!
! Assumes that the determinants are in psi_det
!
! istart, iend, ishift, istep are used in ZMQ parallelization.
!
! if do_right == True then you compute H_TC |Psi>, else H_TC^T |Psi>
END_DOC
integer, intent(in) :: N_st,sze
double precision, intent(inout) :: v_0(sze,N_st), u_0(sze,N_st), s_0(sze,N_st)
logical, intent(in) :: do_right
integer :: k
double precision, allocatable :: u_t(:,:), v_t(:,:), s_t(:,:)
!DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: u_t
allocate(u_t(N_st,N_det),v_t(N_st,N_det),s_t(N_st,N_det))
do k=1,N_st
call dset_order(u_0(1,k),psi_bilinear_matrix_order,N_det)
enddo
v_t = 0.d0
s_t = 0.d0
call dtranspose( &
u_0, &
size(u_0, 1), &
u_t, &
size(u_t, 1), &
N_det, N_st)
call H_tc_s2_u_0_nstates_openmp_work(v_t,s_t,u_t,N_st,sze,1,N_det,0,1, do_right)
deallocate(u_t)
call dtranspose( &
v_t, &
size(v_t, 1), &
v_0, &
size(v_0, 1), &
N_st, N_det)
call dtranspose( &
s_t, &
size(s_t, 1), &
s_0, &
size(s_0, 1), &
N_st, N_det)
deallocate(v_t,s_t)
do k=1,N_st
call dset_order(v_0(1,k),psi_bilinear_matrix_order_reverse,N_det)
call dset_order(s_0(1,k),psi_bilinear_matrix_order_reverse,N_det)
call dset_order(u_0(1,k),psi_bilinear_matrix_order_reverse,N_det)
enddo
end
subroutine H_tc_s2_u_0_nstates_openmp_work(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep, do_right)
use bitmasks
implicit none
BEGIN_DOC
! Computes $v_t = H | u_t\rangle$
!
! Default should be 1,N_det,0,1
!
! if do_right == True then you compute H_TC |Psi>, else H_TC^T |Psi>
END_DOC
integer, intent(in) :: N_st,sze,istart,iend,ishift,istep
double precision, intent(in) :: u_t(N_st,N_det)
logical, intent(in) :: do_right
double precision, intent(out) :: v_t(N_st,sze), s_t(N_st,sze)
PROVIDE ref_bitmask_energy N_int
select case (N_int)
case (1)
call H_tc_s2_u_0_nstates_openmp_work_1(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep,do_right)
case (2)
call H_tc_s2_u_0_nstates_openmp_work_2(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep,do_right)
case (3)
call H_tc_s2_u_0_nstates_openmp_work_3(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep,do_right)
case (4)
call H_tc_s2_u_0_nstates_openmp_work_4(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep,do_right)
case default
call H_tc_s2_u_0_nstates_openmp_work_N_int(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep,do_right)
end select
end
BEGIN_TEMPLATE
subroutine H_tc_s2_u_0_nstates_openmp_work_$N_int(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep,do_right)
use bitmasks
implicit none
BEGIN_DOC
! Computes $v_t = H | u_t \\rangle$ and $s_t = S^2 | u_t\\rangle$
!
! Default should be 1,N_det,0,1
!
! if do_right == True then you compute H_TC |Psi>, else H_TC^T |Psi>
END_DOC
integer, intent(in) :: N_st,sze,istart,iend,ishift,istep
double precision, intent(in) :: u_t(N_st,N_det)
logical, intent(in) :: do_right
double precision, intent(out) :: v_t(N_st,sze), s_t(N_st,sze)
double precision :: hij, sij
integer :: i,j,k,l,kk
integer :: k_a, k_b, l_a, l_b, m_a, m_b
integer :: istate
integer :: krow, kcol, krow_b, kcol_b
integer :: lrow, lcol
integer :: mrow, mcol
integer(bit_kind) :: spindet($N_int)
integer(bit_kind) :: tmp_det($N_int,2)
integer(bit_kind) :: tmp_det2($N_int,2)
integer(bit_kind) :: tmp_det3($N_int,2)
integer(bit_kind), allocatable :: buffer(:,:)
integer :: n_doubles
integer, allocatable :: doubles(:)
integer, allocatable :: singles_a(:)
integer, allocatable :: singles_b(:)
integer, allocatable :: idx(:), idx0(:)
integer :: maxab, n_singles_a, n_singles_b, kcol_prev
integer*8 :: k8
logical :: compute_singles
integer*8 :: last_found, left, right, right_max
double precision :: rss, mem, ratio
double precision, allocatable :: utl(:,:)
integer, parameter :: block_size=128
logical :: u_is_sparse
! call resident_memory(rss)
! mem = dble(singles_beta_csc_size) / 1024.d0**3
!
! compute_singles = (mem+rss > qp_max_mem)
!
! if (.not.compute_singles) then
! provide singles_beta_csc
! endif
compute_singles=.True.
maxab = max(N_det_alpha_unique, N_det_beta_unique)+1
allocate(idx0(maxab))
do i=1,maxab
idx0(i) = i
enddo
! Prepare the array of all alpha single excitations
! -------------------------------------------------
PROVIDE N_int nthreads_davidson
!$OMP PARALLEL DEFAULT(SHARED) NUM_THREADS(nthreads_davidson) &
!$OMP SHARED(psi_bilinear_matrix_rows, N_det, &
!$OMP psi_bilinear_matrix_columns, &
!$OMP psi_det_alpha_unique, psi_det_beta_unique, &
!$OMP n_det_alpha_unique, n_det_beta_unique, N_int, &
!$OMP psi_bilinear_matrix_transp_rows, &
!$OMP psi_bilinear_matrix_transp_columns, &
!$OMP psi_bilinear_matrix_transp_order, N_st, &
!$OMP psi_bilinear_matrix_order_transp_reverse, &
!$OMP psi_bilinear_matrix_columns_loc, &
!$OMP psi_bilinear_matrix_transp_rows_loc, &
!$OMP istart, iend, istep, irp_here, v_t, s_t, &
!$OMP ishift, idx0, u_t, maxab, compute_singles, &
!$OMP singles_alpha_csc,singles_alpha_csc_idx, &
!$OMP singles_beta_csc,singles_beta_csc_idx) &
!$OMP PRIVATE(krow, kcol, tmp_det, spindet, k_a, k_b, i, &
!$OMP lcol, lrow, l_a, l_b, utl, kk, u_is_sparse, &
!$OMP buffer, doubles, n_doubles, umax, &
!$OMP tmp_det2, hij, sij, idx, l, kcol_prev,hmono, htwoe, hthree, &
!$OMP singles_a, n_singles_a, singles_b, ratio, &
!$OMP n_singles_b, k8, last_found,left,right,right_max)
! Alpha/Beta double excitations
! =============================
allocate( buffer($N_int,maxab), &
singles_a(maxab), &
singles_b(maxab), &
doubles(maxab), &
idx(maxab), utl(N_st,block_size))
kcol_prev=-1
! Check if u has multiple zeros
kk=1 ! Avoid division by zero
!$OMP DO
do k=1,N_det
umax = 0.d0
do l=1,N_st
umax = max(umax, dabs(u_t(l,k)))
enddo
if (umax < 1.d-20) then
!$OMP ATOMIC
kk = kk+1
endif
enddo
!$OMP END DO
u_is_sparse = N_det / kk < 20 ! 5%
ASSERT (iend <= N_det)
ASSERT (istart > 0)
ASSERT (istep > 0)
!$OMP DO SCHEDULE(guided,64)
do k_a=istart+ishift,iend,istep ! Loop over all determinants (/!\ not in psidet order)
krow = psi_bilinear_matrix_rows(k_a) ! Index of alpha part of determinant k_a
ASSERT (krow <= N_det_alpha_unique)
kcol = psi_bilinear_matrix_columns(k_a) ! Index of beta part of determinant k_a
ASSERT (kcol <= N_det_beta_unique)
tmp_det(1:$N_int,1) = psi_det_alpha_unique(1:$N_int, krow)
if (kcol /= kcol_prev) then
tmp_det(1:$N_int,2) = psi_det_beta_unique (1:$N_int, kcol)
if (compute_singles) then
call get_all_spin_singles_$N_int( &
psi_det_beta_unique, idx0, &
tmp_det(1,2), N_det_beta_unique, &
singles_b, n_singles_b)
else
n_singles_b = 0
!DIR$ LOOP COUNT avg(1000)
do k8=singles_beta_csc_idx(kcol),singles_beta_csc_idx(kcol+1)-1
n_singles_b = n_singles_b+1
singles_b(n_singles_b) = singles_beta_csc(k8)
enddo
endif
endif
kcol_prev = kcol
! -> Here, tmp_det is determinant k_a
! Loop over singly excited beta columns
! -------------------------------------
!DIR$ LOOP COUNT avg(1000)
do i=1,n_singles_b
lcol = singles_b(i)
tmp_det2(1:$N_int,2) = psi_det_beta_unique(1:$N_int, lcol)
! tmp_det2 is a single excitation of tmp_det in the beta spin
! the alpha part is not defined yet
!---
! if (compute_singles) then
l_a = psi_bilinear_matrix_columns_loc(lcol)
ASSERT (l_a <= N_det)
! rows : | 1 2 3 4 | 1 3 4 6 | .... | 1 2 4 5 |
! cols : | 1 1 1 1 | 2 2 2 2 | .... | 8 8 8 8 |
! index : | 1 2 3 4 | 5 6 7 8 | .... | 58 59 60 61 |
! ^ ^
! | |
! l_a N_det
! l_a is the index in the big vector os size Ndet of the position of the first element of column lcol
! Below we identify all the determinants with the same beta part
!DIR$ UNROLL(8)
!DIR$ LOOP COUNT avg(50000)
do j=1,psi_bilinear_matrix_columns_loc(lcol+1) - psi_bilinear_matrix_columns_loc(lcol)
lrow = psi_bilinear_matrix_rows(l_a)
ASSERT (lrow <= N_det_alpha_unique)
buffer(1:$N_int,j) = psi_det_alpha_unique(1:$N_int, lrow) ! hot spot
ASSERT (l_a <= N_det)
idx(j) = l_a
l_a = l_a+1
enddo
j = j-1
! Get all single excitations from tmp_det(1,1) to buffer(1,?)
call get_all_spin_singles_$N_int( &
buffer, idx, tmp_det(1,1), j, &
singles_a, n_singles_a )
! Loop over alpha singles
! -----------------------
double precision :: umax
!DIR$ LOOP COUNT avg(1000)
do k = 1,n_singles_a,block_size
umax = 0.d0
! Prefetch u_t(:,l_a)
if (u_is_sparse) then
do kk=0,block_size-1
if (k+kk > n_singles_a) exit
l_a = singles_a(k+kk)
ASSERT (l_a <= N_det)
do l=1,N_st
utl(l,kk+1) = u_t(l,l_a)
umax = max(umax, dabs(utl(l,kk+1)))
enddo
enddo
else
do kk=0,block_size-1
if (k+kk > n_singles_a) exit
l_a = singles_a(k+kk)
ASSERT (l_a <= N_det)
utl(:,kk+1) = u_t(:,l_a)
enddo
umax = 1.d0
endif
if (umax < 1.d-20) cycle
do kk=0,block_size-1
if (k+kk > n_singles_a) exit
l_a = singles_a(k+kk)
lrow = psi_bilinear_matrix_rows(l_a)
ASSERT (lrow <= N_det_alpha_unique)
tmp_det2(1:$N_int,1) = psi_det_alpha_unique(1:$N_int, lrow)
! call i_H_j( tmp_det, tmp_det2, $N_int, hij) ! double alpha-beta
if(do_right)then
call htilde_mu_mat_opt_bi_ortho_tot(tmp_det,tmp_det2,$N_int,hij)
else
call htilde_mu_mat_opt_bi_ortho_tot(tmp_det2,tmp_det,$N_int,hij)
endif
call get_s2(tmp_det,tmp_det2,$N_int,sij)
!DIR$ LOOP COUNT AVG(4)
do l=1,N_st
v_t(l,k_a) = v_t(l,k_a) + hij * utl(l,kk+1)
s_t(l,k_a) = s_t(l,k_a) + sij * utl(l,kk+1)
enddo
enddo
enddo
enddo
enddo
!$OMP END DO
!$OMP DO SCHEDULE(guided,64)
do k_a=istart+ishift,iend,istep
! Single and double alpha excitations
! ===================================
! Initial determinant is at k_a in alpha-major representation
! -----------------------------------------------------------------------
krow = psi_bilinear_matrix_rows(k_a)
ASSERT (krow <= N_det_alpha_unique)
kcol = psi_bilinear_matrix_columns(k_a)
ASSERT (kcol <= N_det_beta_unique)
tmp_det(1:$N_int,1) = psi_det_alpha_unique(1:$N_int, krow)
tmp_det(1:$N_int,2) = psi_det_beta_unique (1:$N_int, kcol)
! Initial determinant is at k_b in beta-major representation
! ----------------------------------------------------------------------
k_b = psi_bilinear_matrix_order_transp_reverse(k_a)
ASSERT (k_b <= N_det)
spindet(1:$N_int) = tmp_det(1:$N_int,1)
! Loop inside the beta column to gather all the connected alphas
lcol = psi_bilinear_matrix_columns(k_a)
l_a = psi_bilinear_matrix_columns_loc(lcol)
!DIR$ LOOP COUNT avg(200000)
do i=1,N_det_alpha_unique
if (l_a > N_det) exit
lcol = psi_bilinear_matrix_columns(l_a)
if (lcol /= kcol) exit
lrow = psi_bilinear_matrix_rows(l_a)
ASSERT (lrow <= N_det_alpha_unique)
buffer(1:$N_int,i) = psi_det_alpha_unique(1:$N_int, lrow) ! Hot spot
idx(i) = l_a
l_a = l_a+1
enddo
i = i-1
call get_all_spin_singles_and_doubles_$N_int( &
buffer, idx, spindet, i, &
singles_a, doubles, n_singles_a, n_doubles )
! Compute Hij for all alpha singles
! ----------------------------------
tmp_det2(1:$N_int,2) = psi_det_beta_unique (1:$N_int, kcol)
!DIR$ LOOP COUNT avg(1000)
do i=1,n_singles_a,block_size
umax = 0.d0
! Prefetch u_t(:,l_a)
if (u_is_sparse) then
do kk=0,block_size-1
if (i+kk > n_singles_a) exit
l_a = singles_a(i+kk)
ASSERT (l_a <= N_det)
do l=1,N_st
utl(l,kk+1) = u_t(l,l_a)
umax = max(umax, dabs(utl(l,kk+1)))
enddo
enddo
else
do kk=0,block_size-1
if (i+kk > n_singles_a) exit
l_a = singles_a(i+kk)
ASSERT (l_a <= N_det)
utl(:,kk+1) = u_t(:,l_a)
enddo
umax = 1.d0
endif
if (umax < 1.d-20) cycle
do kk=0,block_size-1
if (i+kk > n_singles_a) exit
l_a = singles_a(i+kk)
lrow = psi_bilinear_matrix_rows(l_a)
ASSERT (lrow <= N_det_alpha_unique)
tmp_det2(1:$N_int,1) = psi_det_alpha_unique(1:$N_int, lrow)
! call i_h_j_single_spin( tmp_det, tmp_det2, $N_int, 1, hij)
if(do_right)then
call htilde_mu_mat_opt_bi_ortho_tot(tmp_det,tmp_det2,$N_int,hij)
else
call htilde_mu_mat_opt_bi_ortho_tot(tmp_det2,tmp_det,$N_int,hij)
endif
!DIR$ LOOP COUNT AVG(4)
do l=1,N_st
v_t(l,k_a) = v_t(l,k_a) + hij * utl(l,kk+1)
enddo
enddo
enddo
! Compute Hij for all alpha doubles
! ----------------------------------
!DIR$ LOOP COUNT avg(50000)
do i=1,n_doubles,block_size
umax = 0.d0
! Prefetch u_t(:,l_a)
if (u_is_sparse) then
do kk=0,block_size-1
if (i+kk > n_doubles) exit
l_a = doubles(i+kk)
ASSERT (l_a <= N_det)
do l=1,N_st
utl(l,kk+1) = u_t(l,l_a)
umax = max(umax, dabs(utl(l,kk+1)))
enddo
enddo
else
do kk=0,block_size-1
if (i+kk > n_doubles) exit
l_a = doubles(i+kk)
ASSERT (l_a <= N_det)
utl(:,kk+1) = u_t(:,l_a)
enddo
umax = 1.d0
endif
if (umax < 1.d-20) cycle
do kk=0,block_size-1
if (i+kk > n_doubles) exit
l_a = doubles(i+kk)
lrow = psi_bilinear_matrix_rows(l_a)
ASSERT (lrow <= N_det_alpha_unique)
tmp_det2(1:N_int,1) = psi_det_alpha_unique(1:N_int, lrow)
! call i_H_j( tmp_det, tmp_det2, $N_int, hij)
! call i_H_j_double_spin( tmp_det(1,1), psi_det_alpha_unique(1, lrow), $N_int, hij)
if(do_right)then
call htilde_mu_mat_opt_bi_ortho_tot(tmp_det,tmp_det2,$N_int,hij)
else
call htilde_mu_mat_opt_bi_ortho_tot(tmp_det2,tmp_det,$N_int,hij)
endif
!DIR$ LOOP COUNT AVG(4)
do l=1,N_st
v_t(l,k_a) = v_t(l,k_a) + hij * utl(l,kk+1)
enddo
enddo
enddo
! Single and double beta excitations
! ==================================
! Initial determinant is at k_a in alpha-major representation
! -----------------------------------------------------------------------
krow = psi_bilinear_matrix_rows(k_a)
kcol = psi_bilinear_matrix_columns(k_a)
tmp_det(1:$N_int,1) = psi_det_alpha_unique(1:$N_int, krow)
tmp_det(1:$N_int,2) = psi_det_beta_unique (1:$N_int, kcol)
spindet(1:$N_int) = tmp_det(1:$N_int,2)
! Initial determinant is at k_b in beta-major representation
! -----------------------------------------------------------------------
k_b = psi_bilinear_matrix_order_transp_reverse(k_a)
ASSERT (k_b <= N_det)
! Loop inside the alpha row to gather all the connected betas
lrow = psi_bilinear_matrix_transp_rows(k_b)
l_b = psi_bilinear_matrix_transp_rows_loc(lrow)
!DIR$ LOOP COUNT avg(200000)
do i=1,N_det_beta_unique
if (l_b > N_det) exit
lrow = psi_bilinear_matrix_transp_rows(l_b)
if (lrow /= krow) exit
lcol = psi_bilinear_matrix_transp_columns(l_b)
ASSERT (lcol <= N_det_beta_unique)
buffer(1:$N_int,i) = psi_det_beta_unique(1:$N_int, lcol)
idx(i) = l_b
l_b = l_b+1
enddo
i = i-1
call get_all_spin_singles_and_doubles_$N_int( &
buffer, idx, spindet, i, &
singles_b, doubles, n_singles_b, n_doubles )
! Compute Hij for all beta singles
! ----------------------------------
tmp_det2(1:$N_int,1) = psi_det_alpha_unique(1:$N_int, krow)
!DIR$ LOOP COUNT avg(1000)
do i=1,n_singles_b,block_size
umax = 0.d0
if (u_is_sparse) then
do kk=0,block_size-1
if (i+kk > n_singles_b) exit
l_b = singles_b(i+kk)
l_a = psi_bilinear_matrix_transp_order(l_b)
ASSERT (l_b <= N_det)
ASSERT (l_a <= N_det)
do l=1,N_st
utl(l,kk+1) = u_t(l,l_a)
umax = max(umax, dabs(utl(l,kk+1)))
enddo
enddo
else
do kk=0,block_size-1
if (i+kk > n_singles_b) exit
l_b = singles_b(i+kk)
l_a = psi_bilinear_matrix_transp_order(l_b)
ASSERT (l_b <= N_det)
ASSERT (l_a <= N_det)
utl(:,kk+1) = u_t(:,l_a)
enddo
umax = 1.d0
endif
if (umax < 1.d-20) cycle
do kk=0,block_size-1
if (i+kk > n_singles_b) exit
l_b = singles_b(i+kk)
l_a = psi_bilinear_matrix_transp_order(l_b)
lcol = psi_bilinear_matrix_transp_columns(l_b)
ASSERT (lcol <= N_det_beta_unique)
tmp_det2(1:$N_int,2) = psi_det_beta_unique (1:$N_int, lcol)
! call i_H_j_single_spin( tmp_det, tmp_det2, $N_int, 2, hij)
if(do_right)then
call htilde_mu_mat_opt_bi_ortho_tot(tmp_det,tmp_det2,$N_int,hij)
else
call htilde_mu_mat_opt_bi_ortho_tot(tmp_det2,tmp_det,$N_int,hij)
endif
!DIR$ LOOP COUNT AVG(4)
do l=1,N_st
v_t(l,k_a) = v_t(l,k_a) + hij * utl(l,kk+1)
enddo
enddo
enddo
! Compute Hij for all beta doubles
! ----------------------------------
!DIR$ LOOP COUNT avg(50000)
do i=1,n_doubles,block_size
umax = 0.d0
if (u_is_sparse) then
do kk=0,block_size-1
if (i+kk > n_doubles) exit
l_b = doubles(i+kk)
l_a = psi_bilinear_matrix_transp_order(l_b)
ASSERT (l_b <= N_det)
ASSERT (l_a <= N_det)
do l=1,N_st
utl(l,kk+1) = u_t(l,l_a)
umax = max(umax, dabs(utl(l,kk+1)))
enddo
enddo
else
do kk=0,block_size-1
if (i+kk > n_doubles) exit
l_b = doubles(i+kk)
l_a = psi_bilinear_matrix_transp_order(l_b)
ASSERT (l_b <= N_det)
ASSERT (l_a <= N_det)
utl(:,kk+1) = u_t(:,l_a)
enddo
umax = 1.d0
endif
if (umax < 1.d-20) cycle
do kk=0,block_size-1
if (i+kk > n_doubles) exit
l_b = doubles(i+kk)
l_a = psi_bilinear_matrix_transp_order(l_b)
lcol = psi_bilinear_matrix_transp_columns(l_b)
ASSERT (lcol <= N_det_beta_unique)
tmp_det2(1:N_int,2) = psi_det_beta_unique(1:N_int, lcol)
! call i_H_j( tmp_det, tmp_det2, $N_int, hij)
! call i_H_j_double_spin( tmp_det(1,2), psi_det_beta_unique(1, lcol), $N_int, hij)
if(do_right)then
call htilde_mu_mat_opt_bi_ortho_tot(tmp_det,tmp_det2,$N_int,hij)
else
call htilde_mu_mat_opt_bi_ortho_tot(tmp_det2,tmp_det,$N_int,hij)
endif
!DIR$ LOOP COUNT AVG(4)
do l=1,N_st
v_t(l,k_a) = v_t(l,k_a) + hij * utl(l,kk+1)
enddo
enddo
enddo
! Diagonal contribution
! =====================
! Initial determinant is at k_a in alpha-major representation
! -----------------------------------------------------------------------
if (u_is_sparse) then
umax = 0.d0
do l=1,N_st
umax = max(umax, dabs(u_t(l,k_a)))
enddo
else
umax = 1.d0
endif
if (umax < 1.d-20) cycle
krow = psi_bilinear_matrix_rows(k_a)
ASSERT (krow <= N_det_alpha_unique)
kcol = psi_bilinear_matrix_columns(k_a)
ASSERT (kcol <= N_det_beta_unique)
tmp_det(1:$N_int,1) = psi_det_alpha_unique(1:$N_int, krow)
tmp_det(1:$N_int,2) = psi_det_beta_unique (1:$N_int, kcol)
double precision, external :: diag_H_mat_elem
double precision :: hmono, htwoe, hthree
! hij = diag_H_mat_elem(tmp_det,$N_int)
call diag_htilde_mu_mat_fock_bi_ortho ($N_int, tmp_det, hmono, htwoe, hthree, hij)
call get_s2(tmp_det,tmp_det,$N_int,sij)
!DIR$ LOOP COUNT AVG(4)
do l=1,N_st
v_t(l,k_a) = v_t(l,k_a) + hij * u_t(l,k_a)
s_t(l,k_a) = s_t(l,k_a) + sij * u_t(l,k_a)
enddo
end do
!$OMP END DO
deallocate(buffer, singles_a, singles_b, doubles, idx, utl)
!$OMP END PARALLEL
end
SUBST [ N_int ]
1;;
2;;
3;;
4;;
N_int;;
END_TEMPLATE

3
src/tc_bi_ortho/u0_h_u0.irp.f → src/tc_bi_ortho/h_tc_u0.irp.f
View File

@ -93,9 +93,6 @@ subroutine H_tc_u_0_nstates_openmp(v_0,u_0,N_st,sze, do_right)
double precision, allocatable :: u_t(:,:), v_t(:,:)
!DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: u_t
allocate(u_t(N_st,N_det),v_t(N_st,N_det))
! provide mo_bi_ortho_tc_one_e mo_bi_ortho_tc_two_e
! provide ref_tc_energy_tot fock_op_2_e_tc_closed_shell
! provide eff_2_e_from_3_e_ab eff_2_e_from_3_e_aa eff_2_e_from_3_e_bb
do k=1,N_st
call dset_order(u_0(1,k),psi_bilinear_matrix_order,N_det)
enddo

6
src/tc_bi_ortho/save_lr_bi_ortho_states.irp.f → src/tc_bi_ortho/print_tc_energy.irp.f
View File

@ -1,4 +1,4 @@
program tc_bi_ortho
program print_tc_energy
implicit none
BEGIN_DOC
! TODO : Put the documentation of the program here
@ -10,6 +10,6 @@ program tc_bi_ortho
read_wf = .True.
touch read_wf
touch my_grid_becke my_n_pt_r_grid my_n_pt_a_grid
call routine_save_left_right_bi_ortho
! call test
call write_tc_energy
end

4
src/tc_bi_ortho/psi_left_qmc.irp.f
View File

@ -17,6 +17,8 @@ BEGIN_PROVIDER [ double precision, psi_bitcleft_bilinear_matrix_values, (N_det,
implicit none
integer :: k, l
!print *, ' providing psi_bitcleft_bilinear_matrix_values'
if(N_det .eq. 1) then
do l = 1, N_states
@ -38,6 +40,8 @@ BEGIN_PROVIDER [ double precision, psi_bitcleft_bilinear_matrix_values, (N_det,
endif
!print *, ' psi_bitcleft_bilinear_matrix_values OK'
END_PROVIDER
! ---

101
src/tc_bi_ortho/psi_r_l_prov.irp.f
View File

@ -136,7 +136,7 @@ BEGIN_PROVIDER [ double precision, psi_r_coef_bi_ortho, (psi_det_size,N_states)
END_PROVIDER
subroutine save_tc_wavefunction_general(ndet,nstates,psidet,sze,dim_psicoef,psilcoef,psircoef)
subroutine save_tc_wavefunction_general(ndet, nstates, psidet, sze, dim_psicoef, psilcoef, psircoef)
implicit none
BEGIN_DOC
! Save the wave function into the |EZFIO| file
@ -192,37 +192,78 @@ subroutine save_tc_wavefunction_general(ndet,nstates,psidet,sze,dim_psicoef,psil
endif
end
subroutine save_tc_bi_ortho_wavefunction
implicit none
if(save_sorted_tc_wf)then
call save_tc_wavefunction_general(N_det,N_states,psi_det_sorted_tc,size(psi_det_sorted_tc, 3),size(psi_l_coef_sorted_bi_ortho, 1),psi_l_coef_sorted_bi_ortho,psi_r_coef_sorted_bi_ortho)
else
call save_tc_wavefunction_general(N_det,N_states,psi_det,size(psi_det, 3), size(psi_l_coef_bi_ortho, 1),psi_l_coef_bi_ortho,psi_r_coef_bi_ortho)
endif
call routine_save_right_bi_ortho
! ---
subroutine save_tc_bi_ortho_wavefunction()
implicit none
if(save_sorted_tc_wf) then
call save_tc_wavefunction_general( N_det, N_states, psi_det_sorted_tc, size(psi_det_sorted_tc, 3) &
, size(psi_l_coef_sorted_bi_ortho, 1), psi_l_coef_sorted_bi_ortho, psi_r_coef_sorted_bi_ortho)
call routine_save_right_sorted_bi_ortho()
else
call save_tc_wavefunction_general( N_det, N_states, psi_det, size(psi_det, 3) &
, size(psi_l_coef_bi_ortho, 1), psi_l_coef_bi_ortho, psi_r_coef_bi_ortho )
call routine_save_right_bi_ortho()
endif
end
subroutine routine_save_right_bi_ortho
implicit none
double precision, allocatable :: coef_tmp(:,:)
integer :: i
allocate(coef_tmp(N_det, N_states))
do i = 1, N_det
coef_tmp(i,1:N_states) = psi_r_coef_sorted_bi_ortho(i,1:N_states)
enddo
call save_wavefunction_general_unormalized(N_det,N_states,psi_det_sorted_tc,size(coef_tmp,1),coef_tmp(1,1))
end
! ---
subroutine routine_save_right_sorted_bi_ortho()
implicit none
integer :: i
double precision, allocatable :: coef_tmp(:,:)
allocate(coef_tmp(N_det, N_states))
do i = 1, N_det
coef_tmp(i,1:N_states) = psi_r_coef_sorted_bi_ortho(i,1:N_states)
enddo
call save_wavefunction_general_unormalized(N_det, N_states, psi_det_sorted_tc, size(coef_tmp, 1), coef_tmp(1,1))
deallocate(coef_tmp)
subroutine routine_save_left_right_bi_ortho
implicit none
double precision, allocatable :: coef_tmp(:,:)
integer :: i,n_states_tmp
n_states_tmp = 2
allocate(coef_tmp(N_det, n_states_tmp))
do i = 1, N_det
coef_tmp(i,1) = psi_r_coef_bi_ortho(i,1)
coef_tmp(i,2) = psi_l_coef_bi_ortho(i,1)
enddo
call save_wavefunction_general_unormalized(N_det,n_states_tmp,psi_det,size(coef_tmp,1),coef_tmp(1,1))
end
subroutine routine_save_left_right_sorted_bi_ortho()
implicit none
integer :: i, n_states_tmp
double precision, allocatable :: coef_tmp(:,:)
n_states_tmp = 2
allocate(coef_tmp(N_det, n_states_tmp))
do i = 1, N_det
coef_tmp(i,1) = psi_r_coef_bi_ortho(i,1)
coef_tmp(i,2) = psi_l_coef_bi_ortho(i,1)
enddo
call save_wavefunction_general_unormalized(N_det, n_states_tmp, psi_det, size(coef_tmp, 1), coef_tmp(1,1))
deallocate(coef_tmp)
end
! ---
subroutine routine_save_right_bi_ortho()
implicit none
integer :: i
double precision, allocatable :: coef_tmp(:,:)
allocate(coef_tmp(N_det, N_states))
do i = 1, N_det
coef_tmp(i,1:N_states) = psi_r_coef_bi_ortho(i,1:N_states)
enddo
call save_wavefunction_general_unormalized(N_det, N_states, psi_det, size(coef_tmp, 1), coef_tmp(1,1))
deallocate(coef_tmp)
end
! ---

21
src/tc_bi_ortho/save_bitcpsileft_for_qmcchem.irp.pouet → src/tc_bi_ortho/save_bitcpsileft_for_qmcchem.irp.f
View File

@ -1,5 +1,18 @@
program save_bitcpsileft_for_qmcchem
implicit none
read_wf = .True.
TOUCH read_wf
call main()
end
subroutine main()
implicit none
integer :: iunit
logical :: exists
double precision :: e_ref
@ -46,7 +59,7 @@ program save_bitcpsileft_for_qmcchem
close(iunit)
end
end subroutine main
! --
@ -61,12 +74,18 @@ subroutine write_lr_spindeterminants()
PROVIDE psi_bitcleft_bilinear_matrix_values
print *, ' saving left determinants'
print *, ' assuming save_for_qmc called before to save right determinants'
print *, ' N_det = ', N_det
print *, ' N_states = ', N_states
allocate(buffer(N_det,N_states))
do l = 1, N_states
do k = 1, N_det
buffer(k,l) = psi_bitcleft_bilinear_matrix_values(k,l)
enddo
enddo
call ezfio_set_spindeterminants_psi_left_coef_matrix_values(buffer)
deallocate(buffer)

89
src/tc_bi_ortho/tc_bi_ortho.irp.f
View File

@ -1,17 +1,25 @@
program tc_bi_ortho
implicit none
BEGIN_DOC
! TODO : Reads psi_det in the EZFIO folder and prints out the left- and right-eigenvectors together with the energy. Saves the left-right wave functions at the end.
!
! TODO : Reads psi_det in the EZFIO folder and prints out the left- and right-eigenvectors together
! with the energy. Saves the left-right wave functions at the end.
!
END_DOC
print *, 'Hello world'
my_grid_becke = .True.
my_n_pt_r_grid = 30
my_n_pt_a_grid = 50
read_wf = .True.
touch read_wf
touch my_grid_becke my_n_pt_r_grid my_n_pt_a_grid
call routine_diag
call save_tc_bi_ortho_wavefunction
touch my_grid_becke my_n_pt_r_grid my_n_pt_a_grid
print*, ' nb of states = ', N_states
print*, ' nb of det = ', N_det
call routine_diag()
call write_tc_energy()
call save_tc_bi_ortho_wavefunction()
end
subroutine test
@ -28,26 +36,53 @@ subroutine test
end
subroutine routine_diag
implicit none
! provide eigval_right_tc_bi_orth
! provide overlap_bi_ortho
! provide htilde_matrix_elmt_bi_ortho
integer ::i,j
print*,'eigval_right_tc_bi_orth = ',eigval_right_tc_bi_orth(1)
print*,'e_tc_left_right = ',e_tc_left_right
print*,'e_tilde_bi_orth_00 = ',e_tilde_bi_orth_00
print*,'e_pt2_tc_bi_orth = ',e_pt2_tc_bi_orth
print*,'e_pt2_tc_bi_orth_single = ',e_pt2_tc_bi_orth_single
print*,'e_pt2_tc_bi_orth_double = ',e_pt2_tc_bi_orth_double
print*,'***'
print*,'e_corr_bi_orth = ',e_corr_bi_orth
print*,'e_corr_bi_orth_proj = ',e_corr_bi_orth_proj
print*,'e_corr_single_bi_orth = ',e_corr_single_bi_orth
print*,'e_corr_double_bi_orth = ',e_corr_double_bi_orth
print*,'Left/right eigenvectors'
do i = 1,N_det
write(*,'(I5,X,(100(F12.7,X)))')i,leigvec_tc_bi_orth(i,1),reigvec_tc_bi_orth(i,1),leigvec_tc_bi_orth(i,1)*reigvec_tc_bi_orth(i,1)
enddo
subroutine routine_diag()
implicit none
integer :: i, j, k
double precision :: dE
! provide eigval_right_tc_bi_orth
! provide overlap_bi_ortho
! provide htilde_matrix_elmt_bi_ortho
if(N_states .eq. 1) then
print*,'eigval_right_tc_bi_orth = ',eigval_right_tc_bi_orth(1)
print*,'e_tc_left_right = ',e_tc_left_right
print*,'e_tilde_bi_orth_00 = ',e_tilde_bi_orth_00
print*,'e_pt2_tc_bi_orth = ',e_pt2_tc_bi_orth
print*,'e_pt2_tc_bi_orth_single = ',e_pt2_tc_bi_orth_single
print*,'e_pt2_tc_bi_orth_double = ',e_pt2_tc_bi_orth_double
print*,'***'
print*,'e_corr_bi_orth = ',e_corr_bi_orth
print*,'e_corr_bi_orth_proj = ',e_corr_bi_orth_proj
print*,'e_corr_single_bi_orth = ',e_corr_single_bi_orth
print*,'e_corr_double_bi_orth = ',e_corr_double_bi_orth
print*,'Left/right eigenvectors'
do i = 1,N_det
write(*,'(I5,X,(100(F12.7,X)))')i,leigvec_tc_bi_orth(i,1),reigvec_tc_bi_orth(i,1),leigvec_tc_bi_orth(i,1)*reigvec_tc_bi_orth(i,1)
enddo
else
print*,'eigval_right_tc_bi_orth : '
do i = 1, N_states
print*, i, eigval_right_tc_bi_orth(i)
enddo
print*,''
print*,'******************************************************'
print*,'TC Excitation energies (au) (eV)'
do i = 2, N_states
dE = eigval_right_tc_bi_orth(i) - eigval_right_tc_bi_orth(1)
print*, i, dE, dE/0.0367502d0
enddo
print*,''
endif
end

395
src/tc_bi_ortho/tc_h_eigvectors.irp.f
View File

@ -25,8 +25,6 @@ subroutine diagonalize_CI_tc
psi_r_coef_bi_ortho(i,j) = reigvec_tc_bi_orth(i,j)
enddo
enddo
! psi_energy(1:N_states) = CI_electronic_energy(1:N_states)
! psi_s2(1:N_states) = CI_s2(1:N_states)
SOFT_TOUCH psi_l_coef_bi_ortho psi_r_coef_bi_ortho
end
@ -37,6 +35,7 @@ end
&BEGIN_PROVIDER [double precision, eigval_left_tc_bi_orth, (N_states)]
&BEGIN_PROVIDER [double precision, reigvec_tc_bi_orth, (N_det,N_states)]
&BEGIN_PROVIDER [double precision, leigvec_tc_bi_orth, (N_det,N_states)]
&BEGIN_PROVIDER [double precision, s2_eigvec_tc_bi_orth, (N_states)]
&BEGIN_PROVIDER [double precision, norm_ground_left_right_bi_orth ]
BEGIN_DOC
@ -44,159 +43,335 @@ end
END_DOC
implicit none
integer :: i, idx_dress, j, istate
integer :: i, idx_dress, j, istate, k
logical :: converged, dagger
integer :: n_real_tc_bi_orth_eigval_right,igood_r,igood_l
double precision, allocatable :: reigvec_tc_bi_orth_tmp(:,:),leigvec_tc_bi_orth_tmp(:,:),eigval_right_tmp(:)
double precision, allocatable :: s2_values_tmp(:), H_prime(:,:), expect_e(:)
double precision, parameter :: alpha = 0.1d0
integer :: i_good_state,i_other_state, i_state
integer, allocatable :: index_good_state_array(:)
logical, allocatable :: good_state_array(:)
double precision, allocatable :: coef_hf_r(:),coef_hf_l(:)
double precision, allocatable :: Stmp(:,:)
integer, allocatable :: iorder(:)
PROVIDE N_det N_int
if(n_det.le.N_det_max_full)then
allocate(reigvec_tc_bi_orth_tmp(N_det,N_det),leigvec_tc_bi_orth_tmp(N_det,N_det),eigval_right_tmp(N_det))
call non_hrmt_real_diag(N_det,htilde_matrix_elmt_bi_ortho,&
leigvec_tc_bi_orth_tmp,reigvec_tc_bi_orth_tmp,&
n_real_tc_bi_orth_eigval_right,eigval_right_tmp)
double precision, allocatable :: coef_hf_r(:),coef_hf_l(:)
integer, allocatable :: iorder(:)
allocate(coef_hf_r(N_det),coef_hf_l(N_det),iorder(N_det))
do i = 1,N_det
iorder(i) = i
coef_hf_r(i) = -dabs(reigvec_tc_bi_orth_tmp(index_HF_psi_det,i))
enddo
call dsort(coef_hf_r,iorder,N_det)
igood_r = iorder(1)
print*,'igood_r, coef_hf_r = ',igood_r,coef_hf_r(1)
do i = 1,N_det
iorder(i) = i
coef_hf_l(i) = -dabs(leigvec_tc_bi_orth_tmp(index_HF_psi_det,i))
enddo
call dsort(coef_hf_l,iorder,N_det)
igood_l = iorder(1)
print*,'igood_l, coef_hf_l = ',igood_l,coef_hf_l(1)
if(n_det .le. N_det_max_full) then
if(igood_r.ne.igood_l.and.igood_r.ne.1)then
print *,''
print *,'Warning, the left and right eigenvectors are "not the same" '
print *,'Warning, the ground state is not dominated by HF...'
print *,'State with largest RIGHT coefficient of HF ',igood_r
print *,'coef of HF in RIGHT eigenvector = ',reigvec_tc_bi_orth_tmp(index_HF_psi_det,igood_r)
print *,'State with largest LEFT coefficient of HF ',igood_l
print *,'coef of HF in LEFT eigenvector = ',leigvec_tc_bi_orth_tmp(index_HF_psi_det,igood_l)
allocate(reigvec_tc_bi_orth_tmp(N_det,N_det),leigvec_tc_bi_orth_tmp(N_det,N_det),eigval_right_tmp(N_det),expect_e(N_det))
allocate (H_prime(N_det,N_det),s2_values_tmp(N_det))
H_prime(1:N_det,1:N_det) = htilde_matrix_elmt_bi_ortho(1:N_det,1:N_det)
if(s2_eig) then
H_prime(1:N_det,1:N_det) += alpha * S2_matrix_all_dets(1:N_det,1:N_det)
do j=1,N_det
H_prime(j,j) = H_prime(j,j) - alpha*expected_s2
enddo
endif
if(state_following_tc)then
print *,'Following the states with the largest coef on HF'
print *,'igood_r,igood_l',igood_r,igood_l
i= igood_r
eigval_right_tc_bi_orth(1) = eigval_right_tmp(i)
do j = 1, N_det
reigvec_tc_bi_orth(j,1) = reigvec_tc_bi_orth_tmp(j,i)
! print*,reigvec_tc_bi_orth(j,1)
enddo
i= igood_l
eigval_left_tc_bi_orth(1) = eigval_right_tmp(i)
do j = 1, N_det
leigvec_tc_bi_orth(j,1) = leigvec_tc_bi_orth_tmp(j,i)
enddo
else
do i = 1, N_states
eigval_right_tc_bi_orth(i) = eigval_right_tmp(i)
eigval_left_tc_bi_orth(i) = eigval_right_tmp(i)
do j = 1, N_det
reigvec_tc_bi_orth(j,i) = reigvec_tc_bi_orth_tmp(j,i)
leigvec_tc_bi_orth(j,i) = leigvec_tc_bi_orth_tmp(j,i)
enddo
enddo
call non_hrmt_real_diag(N_det, H_prime, leigvec_tc_bi_orth_tmp, reigvec_tc_bi_orth_tmp, n_real_tc_bi_orth_eigval_right, eigval_right_tmp)
! do i = 1, N_det
! call get_H_tc_s2_l0_r0(leigvec_tc_bi_orth_tmp(1,i),reigvec_tc_bi_orth_tmp(1,i),1,N_det,expect_e(i), s2_values_tmp(i))
! enddo
call get_H_tc_s2_l0_r0(leigvec_tc_bi_orth_tmp,reigvec_tc_bi_orth_tmp,N_det,N_det,expect_e, s2_values_tmp)
allocate(index_good_state_array(N_det),good_state_array(N_det))
i_state = 0
good_state_array = .False.
if(s2_eig) then
if(only_expected_s2) then
do j = 1, N_det
! Select at least n_states states with S^2 values closed to "expected_s2"
! print*,'s2_values_tmp(j) = ',s2_values_tmp(j),eigval_right_tmp(j),expect_e(j)
if(dabs(s2_values_tmp(j) - expected_s2).le.0.5d0)then
i_state +=1
index_good_state_array(i_state) = j
good_state_array(j) = .True.
endif
if(i_state.eq.N_states) then
exit
endif
enddo
else
do j = 1, N_det
index_good_state_array(j) = j
good_state_array(j) = .True.
enddo
endif
if(i_state .ne. 0) then
! Fill the first "i_state" states that have a correct S^2 value
do j = 1, i_state
do i = 1, N_det
reigvec_tc_bi_orth(i,j) = reigvec_tc_bi_orth_tmp(i,index_good_state_array(j))
leigvec_tc_bi_orth(i,j) = leigvec_tc_bi_orth_tmp(i,index_good_state_array(j))
enddo
eigval_right_tc_bi_orth(j) = expect_e(index_good_state_array(j))
eigval_left_tc_bi_orth(j) = expect_e(index_good_state_array(j))
s2_eigvec_tc_bi_orth(j) = s2_values_tmp(index_good_state_array(j))
enddo
i_other_state = 0
do j = 1, N_det
if(good_state_array(j))cycle
i_other_state +=1
if(i_state+i_other_state.gt.n_states)then
exit
endif
do i = 1, N_det
reigvec_tc_bi_orth(i,i_state+i_other_state) = reigvec_tc_bi_orth_tmp(i,j)
leigvec_tc_bi_orth(i,i_state+i_other_state) = leigvec_tc_bi_orth_tmp(i,j)
enddo
eigval_right_tc_bi_orth(i_state+i_other_state) = eigval_right_tmp(j)
eigval_left_tc_bi_orth (i_state+i_other_state) = eigval_right_tmp(j)
s2_eigvec_tc_bi_orth(i_state+i_other_state) = s2_values_tmp(i_state+i_other_state)
enddo
else ! istate == 0
print*,''
print*,'!!!!!!!! WARNING !!!!!!!!!'
print*,' Within the ',N_det,'determinants selected'
print*,' and the ',N_states_diag,'states requested'
print*,' We did not find only states with S^2 values close to ',expected_s2
print*,' We will then set the first N_states eigenvectors of the H matrix'
print*,' as the CI_eigenvectors'
print*,' You should consider more states and maybe ask for s2_eig to be .True. or just enlarge the CI space'
print*,''
do j = 1, min(N_states_diag, N_det)
do i = 1, N_det
leigvec_tc_bi_orth(i,j) = leigvec_tc_bi_orth_tmp(i,j)
reigvec_tc_bi_orth(i,j) = reigvec_tc_bi_orth_tmp(i,j)
enddo
eigval_right_tc_bi_orth(j) = eigval_right_tmp(j)
eigval_left_tc_bi_orth (j) = eigval_right_tmp(j)
s2_eigvec_tc_bi_orth(j) = s2_values_tmp(j)
enddo
endif ! istate .ne. 0
else ! s2_eig
allocate(coef_hf_r(N_det),coef_hf_l(N_det),iorder(N_det))
do i = 1,N_det
iorder(i) = i
coef_hf_r(i) = -dabs(reigvec_tc_bi_orth_tmp(index_HF_psi_det,i))
enddo
call dsort(coef_hf_r,iorder,N_det)
igood_r = iorder(1)
print*,'igood_r, coef_hf_r = ',igood_r,coef_hf_r(1)
do i = 1,N_det
iorder(i) = i
coef_hf_l(i) = -dabs(leigvec_tc_bi_orth_tmp(index_HF_psi_det,i))
enddo
call dsort(coef_hf_l,iorder,N_det)
igood_l = iorder(1)
print*,'igood_l, coef_hf_l = ',igood_l,coef_hf_l(1)
if(igood_r.ne.igood_l .and. igood_r.ne.1) then
print *,''
print *,'Warning, the left and right eigenvectors are "not the same" '
print *,'Warning, the ground state is not dominated by HF...'
print *,'State with largest RIGHT coefficient of HF ',igood_r
print *,'coef of HF in RIGHT eigenvector = ',reigvec_tc_bi_orth_tmp(index_HF_psi_det,igood_r)
print *,'State with largest LEFT coefficient of HF ',igood_l
print *,'coef of HF in LEFT eigenvector = ',leigvec_tc_bi_orth_tmp(index_HF_psi_det,igood_l)
endif
if(state_following_tc) then
print *,'Following the states with the largest coef on HF'
print *,'igood_r,igood_l',igood_r,igood_l
i = igood_r
eigval_right_tc_bi_orth(1) = eigval_right_tmp(i)
do j = 1, N_det
reigvec_tc_bi_orth(j,1) = reigvec_tc_bi_orth_tmp(j,i)
enddo
i = igood_l
eigval_left_tc_bi_orth(1) = eigval_right_tmp(i)
do j = 1, N_det
leigvec_tc_bi_orth(j,1) = leigvec_tc_bi_orth_tmp(j,i)
enddo
else
do i = 1, N_states
eigval_right_tc_bi_orth(i) = eigval_right_tmp(i)
eigval_left_tc_bi_orth(i) = eigval_right_tmp(i)
do j = 1, N_det
reigvec_tc_bi_orth(j,i) = reigvec_tc_bi_orth_tmp(j,i)
leigvec_tc_bi_orth(j,i) = leigvec_tc_bi_orth_tmp(j,i)
enddo
enddo
endif
endif
else
else ! n_det > N_det_max_full
double precision, allocatable :: H_jj(:),vec_tmp(:,:)
external htc_bi_ortho_calc_tdav
external htcdag_bi_ortho_calc_tdav
external H_tc_u_0_opt
external H_tc_dagger_u_0_opt
external H_tc_s2_dagger_u_0_opt
external H_tc_s2_u_0_opt
allocate(H_jj(N_det),vec_tmp(N_det,n_states_diag))
do i = 1, N_det
call htilde_mu_mat_bi_ortho_tot(psi_det(1,1,i), psi_det(1,1,i), N_int, H_jj(i))
enddo
!!!! Preparing the left-eigenvector
print*,'---------------------------------'
print*,'---------------------------------'
print*,'Computing the left-eigenvector '
print*,'---------------------------------'
print*,'---------------------------------'
!!!! Preparing the left-eigenvector
vec_tmp = 0.d0
do istate = 1, N_states
vec_tmp(1:N_det,istate) = psi_l_coef_bi_ortho(1:N_det,istate)
vec_tmp(1:N_det,istate) = psi_l_coef_bi_ortho(1:N_det,istate)
enddo
do istate = N_states+1, n_states_diag
vec_tmp(istate,istate) = 1.d0
vec_tmp(istate,istate) = 1.d0
enddo
!call davidson_general_ext_rout_nonsym_b1space(vec_tmp, H_jj, eigval_left_tc_bi_orth, N_det, n_states, n_states_diag, converged, htcdag_bi_ortho_calc_tdav)
!call davidson_general_ext_rout_nonsym_b1space(vec_tmp, H_jj, eigval_left_tc_bi_orth, N_det, n_states, n_states_diag, converged, H_tc_dagger_u_0_opt)
integer :: n_it_max,i_it
n_it_max = 1
converged = .False.
i_it = 0
do while (.not.converged)
call davidson_hs2_nonsym_b1space(vec_tmp, H_jj, s2_eigvec_tc_bi_orth, eigval_left_tc_bi_orth, N_det, n_states, n_states_diag, n_it_max, converged, H_tc_s2_dagger_u_0_opt)
i_it += 1
if(i_it .gt. 5) exit
enddo
! call davidson_general_ext_rout_nonsym_b1space(vec_tmp, H_jj, eigval_left_tc_bi_orth, N_det, n_states, n_states_diag, converged, htcdag_bi_ortho_calc_tdav)
call davidson_general_ext_rout_nonsym_b1space(vec_tmp, H_jj, eigval_left_tc_bi_orth, N_det, n_states, n_states_diag, converged, H_tc_dagger_u_0_opt)
do istate = 1, N_states
leigvec_tc_bi_orth(1:N_det,istate) = vec_tmp(1:N_det,istate)
leigvec_tc_bi_orth(1:N_det,istate) = vec_tmp(1:N_det,istate)
enddo
print*,'---------------------------------'
print*,'---------------------------------'
print*,'Computing the right-eigenvector '
!!!! Preparing the right-eigenvector
print*,'---------------------------------'
print*,'---------------------------------'
!!!! Preparing the right-eigenvector
vec_tmp = 0.d0
do istate = 1, N_states
vec_tmp(1:N_det,istate) = psi_r_coef_bi_ortho(1:N_det,istate)
vec_tmp(1:N_det,istate) = psi_r_coef_bi_ortho(1:N_det,istate)
enddo
do istate = N_states+1, n_states_diag
vec_tmp(istate,istate) = 1.d0
vec_tmp(istate,istate) = 1.d0
enddo
!call davidson_general_ext_rout_nonsym_b1space(vec_tmp, H_jj, eigval_right_tc_bi_orth, N_det, n_states, n_states_diag, converged, htc_bi_ortho_calc_tdav)
!call davidson_general_ext_rout_nonsym_b1space(vec_tmp, H_jj, eigval_right_tc_bi_orth, N_det, n_states, n_states_diag, converged, H_tc_u_0_opt)
converged = .False.
i_it = 0
do while (.not. converged)
call davidson_hs2_nonsym_b1space(vec_tmp, H_jj, s2_eigvec_tc_bi_orth, eigval_right_tc_bi_orth, N_det, n_states, n_states_diag, n_it_max, converged, H_tc_s2_u_0_opt)
i_it += 1
if(i_it .gt. 5) exit
enddo
! call davidson_general_ext_rout_nonsym_b1space(vec_tmp, H_jj, eigval_right_tc_bi_orth, N_det, n_states, n_states_diag, converged, htc_bi_ortho_calc_tdav)
call davidson_general_ext_rout_nonsym_b1space(vec_tmp, H_jj, eigval_right_tc_bi_orth, N_det, n_states, n_states_diag, converged, H_tc_u_0_opt)
do istate = 1, N_states
reigvec_tc_bi_orth(1:N_det,istate) = vec_tmp(1:N_det,istate)
reigvec_tc_bi_orth(1:N_det,istate) = vec_tmp(1:N_det,istate)
enddo
deallocate(H_jj)
endif
call bi_normalize(leigvec_tc_bi_orth,reigvec_tc_bi_orth,size(reigvec_tc_bi_orth,1),N_det,N_states)
print*,'leigvec_tc_bi_orth(1,1),reigvec_tc_bi_orth(1,1) = ',leigvec_tc_bi_orth(1,1),reigvec_tc_bi_orth(1,1)
norm_ground_left_right_bi_orth = 0.d0
do j = 1, N_det
norm_ground_left_right_bi_orth += leigvec_tc_bi_orth(j,1) * reigvec_tc_bi_orth(j,1)
enddo
print*,'norm l/r = ',norm_ground_left_right_bi_orth
endif
call bi_normalize(leigvec_tc_bi_orth, reigvec_tc_bi_orth, size(reigvec_tc_bi_orth, 1), N_det, N_states)
! check bi-orthogonality
allocate(Stmp(N_states,N_states))
call dgemm( 'T', 'N', N_states, N_states, N_det, 1.d0 &
, leigvec_tc_bi_orth(1,1), size(leigvec_tc_bi_orth, 1), reigvec_tc_bi_orth(1,1), size(reigvec_tc_bi_orth, 1) &
, 0.d0, Stmp(1,1), size(Stmp, 1) )
print *, ' overlap matrix between states:'
do i = 1, N_states
write(*,'(1000(F16.10,X))') Stmp(i,:)
enddo
deallocate(Stmp)
print*,'leigvec_tc_bi_orth(1,1),reigvec_tc_bi_orth(1,1) = ', leigvec_tc_bi_orth(1,1), reigvec_tc_bi_orth(1,1)
do i = 1, N_states
norm_ground_left_right_bi_orth = 0.d0
do j = 1, N_det
norm_ground_left_right_bi_orth += leigvec_tc_bi_orth(j,i) * reigvec_tc_bi_orth(j,i)
enddo
print*,' state ', i
print*,' norm l/r = ', norm_ground_left_right_bi_orth
print*,' <S2> = ', s2_eigvec_tc_bi_orth(i)
enddo
double precision, allocatable :: buffer(:,:)
allocate(buffer(N_det,N_states))
do k = 1, N_states
do i = 1, N_det
psi_l_coef_bi_ortho(i,k) = leigvec_tc_bi_orth(i,k)
buffer(i,k) = leigvec_tc_bi_orth(i,k)
enddo
enddo
TOUCH psi_l_coef_bi_ortho
call ezfio_set_tc_bi_ortho_psi_l_coef_bi_ortho(buffer)
do k = 1, N_states
do i = 1, N_det
psi_r_coef_bi_ortho(i,k) = reigvec_tc_bi_orth(i,k)
buffer(i,k) = reigvec_tc_bi_orth(i,k)
enddo
enddo
TOUCH psi_r_coef_bi_ortho
call ezfio_set_tc_bi_ortho_psi_r_coef_bi_ortho(buffer)
deallocate(buffer)
END_PROVIDER
subroutine bi_normalize(u_l,u_r,n,ld,nstates)
subroutine bi_normalize(u_l, u_r, n, ld, nstates)
BEGIN_DOC
!!!! Normalization of the scalar product of the left/right eigenvectors
END_DOC
implicit none
integer, intent(in) :: n, ld, nstates
double precision, intent(inout) :: u_l(ld,nstates), u_r(ld,nstates)
integer, intent(in) :: n,ld,nstates
integer :: i
double precision :: accu, tmp
integer :: i, j
double precision :: accu, tmp
do i = 1, nstates
!!!! Normalization of right eigenvectors |Phi>
accu = 0.d0
do j = 1, n
accu += u_r(j,i) * u_r(j,i)
enddo
accu = 1.d0/dsqrt(accu)
print*,'accu_r = ',accu
do j = 1, n
u_r(j,i) *= accu
enddo
tmp = u_r(1,i) / dabs(u_r(1,i))
do j = 1, n
u_r(j,i) *= tmp
enddo
!!!! Adaptation of the norm of the left eigenvector such that <chi|Phi> = 1
accu = 0.d0
do j = 1, n
accu += u_l(j,i) * u_r(j,i)
! print*,j, u_l(j,i) , u_r(j,i)
enddo
if(accu.gt.0.d0)then
!!!! Normalization of right eigenvectors |Phi>
accu = 0.d0
do j = 1, n
accu += u_r(j,i) * u_r(j,i)
enddo
accu = 1.d0/dsqrt(accu)
else
accu = 1.d0/dsqrt(-accu)
endif
tmp = (u_l(1,i) * u_r(1,i) )/dabs(u_l(1,i) * u_r(1,i))
do j = 1, n
u_l(j,i) *= accu * tmp
u_r(j,i) *= accu
enddo
print*,'accu_r = ',accu
do j = 1, n
u_r(j,i) *= accu
enddo
tmp = u_r(1,i) / dabs(u_r(1,i))
do j = 1, n
u_r(j,i) *= tmp
enddo
!!!! Adaptation of the norm of the left eigenvector such that <chi|Phi> = 1
accu = 0.d0
do j = 1, n
accu += u_l(j,i) * u_r(j,i)
!print*,j, u_l(j,i) , u_r(j,i)
enddo
print*,'accu_lr = ', accu
if(accu.gt.0.d0)then
accu = 1.d0/dsqrt(accu)
else
accu = 1.d0/dsqrt(-accu)
endif
tmp = (u_l(1,i) * u_r(1,i) )/dabs(u_l(1,i) * u_r(1,i))
do j = 1, n
u_l(j,i) *= accu * tmp
u_r(j,i) *= accu
enddo
enddo
end

5
src/tc_bi_ortho/tc_hmat.irp.f
View File

@ -12,6 +12,11 @@
double precision :: hmono,htwoe,hthree,htot
PROVIDE N_int
i = 1
j = 1
call htilde_mu_mat_bi_ortho(psi_det(1,1,j), psi_det(1,1,i), N_int, hmono, htwoe, hthree, htot)
!$OMP PARALLEL DO SCHEDULE(GUIDED) DEFAULT(NONE) PRIVATE(i,j,hmono, htwoe, hthree, htot) &
!$OMP SHARED (N_det, psi_det, N_int,htilde_matrix_elmt_bi_ortho)
do i = 1, N_det

34
src/tc_bi_ortho/tc_utils.irp.f Normal file
View File

@ -0,0 +1,34 @@
subroutine write_tc_energy()
implicit none
integer :: i, j, k
double precision :: hmono, htwoe, hthree, htot
double precision :: E_TC, O_TC
do k = 1, n_states
E_TC = 0.d0
do i = 1, N_det
do j = 1, N_det
!htot = htilde_matrix_elmt_bi_ortho(i,j)
call htilde_mu_mat_bi_ortho(psi_det(1,1,i), psi_det(1,1,j), N_int, hmono, htwoe, hthree, htot)
E_TC = E_TC + psi_l_coef_bi_ortho(i,k) * psi_r_coef_bi_ortho(j,k) * htot
!E_TC = E_TC + leigvec_tc_bi_orth(i,k) * reigvec_tc_bi_orth(j,k) * htot
enddo
enddo
O_TC = 0.d0
do i = 1, N_det
!O_TC = O_TC + leigvec_tc_bi_orth(i,k) * reigvec_tc_bi_orth(i,k)
O_TC = O_TC + psi_l_coef_bi_ortho(i,k) * psi_r_coef_bi_ortho(i,k)
enddo
print *, ' state :', k
print *, " E_TC = ", E_TC / O_TC
print *, " O_TC = ", O_TC
enddo
end

162
src/tc_bi_ortho/test_s2_tc.irp.f Normal file
View File

@ -0,0 +1,162 @@
program test_tc
implicit none
read_wf = .True.
my_grid_becke = .True.
my_n_pt_r_grid = 30
my_n_pt_a_grid = 50
read_wf = .True.
touch read_wf
touch my_grid_becke my_n_pt_r_grid my_n_pt_a_grid
call routine_test_s2
call routine_test_s2_davidson
end
subroutine routine_test_s2
implicit none
logical :: do_right
integer :: sze ,i, N_st, j
double precision :: sij, accu_e, accu_s, accu_e_0, accu_s_0
double precision, allocatable :: v_0_ref(:,:),u_0(:,:),s_0_ref(:,:)
double precision, allocatable :: v_0_new(:,:),s_0_new(:,:)
sze = N_det
N_st = 1
allocate(v_0_ref(N_det,1),u_0(N_det,1),s_0_ref(N_det,1),s_0_new(N_det,1),v_0_new(N_det,1))
print*,'Checking first the Left '
do_right = .False.
do i = 1, sze
u_0(i,1) = psi_l_coef_bi_ortho(i,1)
enddo
call H_tc_u_0_nstates_openmp(v_0_ref,u_0,N_st,sze, do_right)
s_0_ref = 0.d0
do i = 1, sze
do j = 1, sze
call get_s2(psi_det(1,1,i),psi_det(1,1,j),N_int,sij)
s_0_ref(i,1) += u_0(j,1) * sij
enddo
enddo
call H_tc_s2_u_0_nstates_openmp(v_0_new,s_0_new,u_0,N_st,sze, do_right)
accu_e = 0.d0
accu_s = 0.d0
accu_e_0 = 0.d0
accu_s_0 = 0.d0
do i = 1, sze
accu_e_0 += v_0_ref(i,1) * psi_r_coef_bi_ortho(i,1)
accu_s_0 += s_0_ref(i,1) * psi_r_coef_bi_ortho(i,1)
accu_e += dabs(v_0_ref(i,1) - v_0_new(i,1))
accu_s += dabs(s_0_ref(i,1) - s_0_new(i,1))
enddo
print*,'accu_e = ',accu_e
print*,'accu_s = ',accu_s
print*,'accu_e_0 = ',accu_e_0
print*,'accu_s_0 = ',accu_s_0
print*,'Checking then the right '
do_right = .True.
do i = 1, sze
u_0(i,1) = psi_r_coef_bi_ortho(i,1)
enddo
call H_tc_u_0_nstates_openmp(v_0_ref,u_0,N_st,sze, do_right)
s_0_ref = 0.d0
do i = 1, sze
do j = 1, sze
call get_s2(psi_det(1,1,i),psi_det(1,1,j),N_int,sij)
s_0_ref(i,1) += u_0(j,1) * sij
enddo
enddo
call H_tc_s2_u_0_nstates_openmp(v_0_new,s_0_new,u_0,N_st,sze, do_right)
accu_e = 0.d0
accu_s = 0.d0
accu_e_0 = 0.d0
accu_s_0 = 0.d0
do i = 1, sze
accu_e_0 += v_0_ref(i,1) * psi_l_coef_bi_ortho(i,1)
accu_s_0 += s_0_ref(i,1) * psi_l_coef_bi_ortho(i,1)
accu_e += dabs(v_0_ref(i,1) - v_0_new(i,1))
accu_s += dabs(s_0_ref(i,1) - s_0_new(i,1))
enddo
print*,'accu_e = ',accu_e
print*,'accu_s = ',accu_s
print*,'accu_e_0 = ',accu_e_0
print*,'accu_s_0 = ',accu_s_0
end
subroutine routine_test_s2_davidson
implicit none
double precision, allocatable :: H_jj(:),vec_tmp(:,:), energies(:) , s2(:)
integer :: i,istate
logical :: converged
external H_tc_s2_dagger_u_0_opt
external H_tc_s2_u_0_opt
allocate(H_jj(N_det),vec_tmp(N_det,n_states_diag),energies(n_states_diag), s2(n_states_diag))
do i = 1, N_det
call htilde_mu_mat_bi_ortho_tot(psi_det(1,1,i), psi_det(1,1,i), N_int, H_jj(i))
enddo
! Preparing the left-eigenvector
print*,'Computing the left-eigenvector '
vec_tmp = 0.d0
do istate = 1, N_states
vec_tmp(1:N_det,istate) = psi_l_coef_bi_ortho(1:N_det,istate)
enddo
do istate = N_states+1, n_states_diag
vec_tmp(istate,istate) = 1.d0
enddo
do istate = 1, N_states
leigvec_tc_bi_orth(1:N_det,istate) = vec_tmp(1:N_det,istate)
enddo
integer :: n_it_max
n_it_max = 1
call davidson_hs2_nonsym_b1space(vec_tmp, H_jj, s2, energies, N_det, n_states, n_states_diag, n_it_max, converged, H_tc_s2_dagger_u_0_opt)
double precision, allocatable :: v_0_new(:,:),s_0_new(:,:)
integer :: sze,N_st
logical :: do_right
sze = N_det
N_st = 1
do_right = .False.
allocate(s_0_new(N_det,1),v_0_new(N_det,1))
call H_tc_s2_u_0_nstates_openmp(v_0_new,s_0_new,vec_tmp,N_st,sze, do_right)
double precision :: accu_e_0, accu_s_0
accu_e_0 = 0.d0
accu_s_0 = 0.d0
do i = 1, sze
accu_e_0 += v_0_new(i,1) * vec_tmp(i,1)
accu_s_0 += s_0_new(i,1) * vec_tmp(i,1)
enddo
print*,'energies = ',energies
print*,'s2 = ',s2
print*,'accu_e_0',accu_e_0
print*,'accu_s_0',accu_s_0
! Preparing the right-eigenvector
print*,'Computing the right-eigenvector '
vec_tmp = 0.d0
do istate = 1, N_states
vec_tmp(1:N_det,istate) = psi_r_coef_bi_ortho(1:N_det,istate)
enddo
do istate = N_states+1, n_states_diag
vec_tmp(istate,istate) = 1.d0
enddo
do istate = 1, N_states
leigvec_tc_bi_orth(1:N_det,istate) = vec_tmp(1:N_det,istate)
enddo
n_it_max = 1
call davidson_hs2_nonsym_b1space(vec_tmp, H_jj, s2, energies, N_det, n_states, n_states_diag, n_it_max, converged, H_tc_s2_u_0_opt)
sze = N_det
N_st = 1
do_right = .True.
v_0_new = 0.d0
s_0_new = 0.d0
call H_tc_s2_u_0_nstates_openmp(v_0_new,s_0_new,vec_tmp,N_st,sze, do_right)
accu_e_0 = 0.d0
accu_s_0 = 0.d0
do i = 1, sze
accu_e_0 += v_0_new(i,1) * vec_tmp(i,1)
accu_s_0 += s_0_new(i,1) * vec_tmp(i,1)
enddo
print*,'energies = ',energies
print*,'s2 = ',s2
print*,'accu_e_0',accu_e_0
print*,'accu_s_0',accu_s_0
end

40
src/tc_keywords/EZFIO.cfg
View File

@ -6,7 +6,7 @@ default: False
[comp_left_eigv]
type: logical
doc: If |true|, computes also the left-eigenvector
doc: If |true|, computes also the left-eigenvector
interface: ezfio,provider,ocaml
default: False
@ -14,7 +14,7 @@ default: False
type: logical
doc: If |true|, three-body terms are included
interface: ezfio,provider,ocaml
default: True
default: True
[pure_three_body_h_tc]
type: logical
@ -30,13 +30,13 @@ default: False
[core_tc_op]
type: logical
doc: If |true|, takes the usual Hamiltonian for core orbitals (assumed to be doubly occupied)
doc: If |true|, takes the usual Hamiltonian for core orbitals (assumed to be doubly occupied)
interface: ezfio,provider,ocaml
default: False
[full_tc_h_solver]
type: logical
doc: If |true|, you diagonalize the full TC H matrix
doc: If |true|, you diagonalize the full TC H matrix
interface: ezfio,provider,ocaml
default: False
@ -46,6 +46,12 @@ doc: Thresholds on the energy for iterative Davidson used in TC
interface: ezfio,provider,ocaml
default: 1.e-5
[thrsh_cycle_tc]
type: Threshold
doc: Thresholds to cycle the integrals with the envelop
interface: ezfio,provider,ocaml
default: 1.e-10
[max_it_dav]
type: integer
doc: nb max of iteration in Davidson used in TC
@ -60,11 +66,11 @@ default: 0.000005
[thresh_psi_r_norm]
type: logical
doc: If |true|, you prune the WF to compute the PT1 coef based on the norm. If False, the pruning is done through the amplitude on the right-coefficient.
doc: If |true|, you prune the WF to compute the PT1 coef based on the norm. If False, the pruning is done through the amplitude on the right-coefficient.
interface: ezfio,provider,ocaml
default: False
[state_following_tc]
[state_following_tc]
type: logical
doc: If |true|, the states are re-ordered to match the input states
default: False
@ -78,7 +84,7 @@ default: True
[symetric_fock_tc]
type: logical
doc: If |true|, using F+F^t as Fock TC
doc: If |true|, using F+F^t as Fock TC
interface: ezfio,provider,ocaml
default: False
@ -126,7 +132,7 @@ default: 1.e-6
[maxovl_tc]
type: logical
doc: If |true|, maximize the overlap between orthogonalized left- and right eigenvectors
doc: If |true|, maximize the overlap between orthogonalized left- and right eigenvectors
interface: ezfio,provider,ocaml
default: False
@ -152,7 +158,7 @@ default: 0.
type: character*(32)
doc: Type of TCSCF algorithm used. Possible choices are [Simple | DIIS]
interface: ezfio,provider,ocaml
default: Simple
default: DIIS
[im_thresh_tcscf]
type: Threshold
@ -180,21 +186,15 @@ default: 1.e-6
[var_tc]
type: logical
doc: If |true|, use VAR-TC
doc: If |true|, use VAR-TC
interface: ezfio,provider,ocaml
default: False
[read_tc_integ]
type: logical
doc: If |true|, read integrals: int2_grad1_u12_ao, tc_grad_square_ao and tc_grad_and_lapl_ao
[io_tc_integ]
type: Disk_access
doc: Read/Write integrals int2_grad1_u12_ao, tc_grad_square_ao and tc_grad_and_lapl_ao from/to disk [ Write | Read | None ]
interface: ezfio,provider,ocaml
default: False
[write_tc_integ]
type: logical
doc: If |true|, write integrals: int2_grad1_u12_ao, tc_grad_square_ao and tc_grad_and_lapl_ao
interface: ezfio,provider,ocaml
default: False
default: None
[debug_tc_pt2]
type: integer

214
src/tc_scf/test_int.irp.f
View File

@ -21,25 +21,22 @@ program test_ints
touch my_extra_grid_becke my_n_pt_r_extra_grid my_n_pt_a_extra_grid
!! OK
!call routine_int2_u_grad1u_j1b2
!! OK
!call routine_v_ij_erf_rk_cst_mu_j1b
!! OK
! call routine_int2_u_grad1u_j1b2
! OK
! call routine_v_ij_erf_rk_cst_mu_j1b
! OK
! call routine_x_v_ij_erf_rk_cst_mu_j1b
!! OK
! call routine_v_ij_u_cst_mu_j1b
!! OK
!call routine_int2_u2_j1b2
!! OK
!call routine_int2_u_grad1u_x_j1b2
!! OK
! OK
! call routine_int2_u2_j1b2
! OK
! call routine_int2_u_grad1u_x_j1b2
! OK
! call routine_int2_grad1u2_grad2u2_j1b2
! call routine_int2_u_grad1u_j1b2
! call test_total_grad_lapl
! call test_total_grad_square
! call test_int2_grad1_u12_ao_test
! call routine_v_ij_u_cst_mu_j1b_test
! call test_ao_tc_int_chemist
! call test_grid_points_ao
! call test_tc_scf
@ -53,12 +50,12 @@ program test_ints
!call test_two_e_tc_non_hermit_integral()
call test_tc_grad_square_ao_test()
PROVIDE TC_HF_energy VARTC_HF_energy
print *, ' TC_HF_energy = ', TC_HF_energy
print *, ' VARTC_HF_energy = ', VARTC_HF_energy
! call test_tc_grad_square_ao_test()
!!PROVIDE TC_HF_energy VARTC_HF_energy
!!print *, ' TC_HF_energy = ', TC_HF_energy
!!print *, ' VARTC_HF_energy = ', VARTC_HF_energy
call test_old_ints
end
! ---
@ -157,6 +154,9 @@ subroutine routine_int2_u_grad1u_j1b2
enddo
enddo
enddo
print*,'******'
print*,'******'
print*,'routine_int2_u_grad1u_j1b2'
print*,'accu_abs = ',accu_abs/dble(ao_num)**4
print*,'accu_relat = ',accu_relat/dble(ao_num)**4
@ -169,20 +169,6 @@ subroutine routine_v_ij_erf_rk_cst_mu_j1b
integer :: i,j,ipoint,k,l
double precision :: weight,accu_relat, accu_abs, contrib
double precision, allocatable :: array(:,:,:,:), array_ref(:,:,:,:)
! print*,'ao_overlap_abs = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_overlap_abs(i,:)
! enddo
! print*,'center = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_prod_center(2,i,:)
! enddo
! print*,'sigma = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_prod_sigma(i,:)
! enddo
allocate(array(ao_num, ao_num, ao_num, ao_num))
array = 0.d0
allocate(array_ref(ao_num, ao_num, ao_num, ao_num))
@ -215,6 +201,9 @@ subroutine routine_v_ij_erf_rk_cst_mu_j1b
enddo
enddo
enddo
print*,'******'
print*,'******'
print*,'routine_v_ij_erf_rk_cst_mu_j1b'
print*,'accu_abs = ',accu_abs/dble(ao_num)**4
print*,'accu_relat = ',accu_relat/dble(ao_num)**4
@ -228,20 +217,6 @@ subroutine routine_x_v_ij_erf_rk_cst_mu_j1b
integer :: i,j,ipoint,k,l,m
double precision :: weight,accu_relat, accu_abs, contrib
double precision, allocatable :: array(:,:,:,:), array_ref(:,:,:,:)
! print*,'ao_overlap_abs = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_overlap_abs(i,:)
! enddo
! print*,'center = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_prod_center(2,i,:)
! enddo
! print*,'sigma = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_prod_sigma(i,:)
! enddo
allocate(array(ao_num, ao_num, ao_num, ao_num))
array = 0.d0
allocate(array_ref(ao_num, ao_num, ao_num, ao_num))
@ -276,6 +251,10 @@ subroutine routine_x_v_ij_erf_rk_cst_mu_j1b
enddo
enddo
enddo
print*,'******'
print*,'******'
print*,'routine_x_v_ij_erf_rk_cst_mu_j1b'
print*,'accu_abs = ',accu_abs/dble(ao_num)**4
print*,'accu_relat = ',accu_relat/dble(ao_num)**4
@ -290,20 +269,6 @@ subroutine routine_v_ij_u_cst_mu_j1b_test
integer :: i,j,ipoint,k,l
double precision :: weight,accu_relat, accu_abs, contrib
double precision, allocatable :: array(:,:,:,:), array_ref(:,:,:,:)
! print*,'ao_overlap_abs = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_overlap_abs(i,:)
! enddo
! print*,'center = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_prod_center(2,i,:)
! enddo
! print*,'sigma = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_prod_sigma(i,:)
! enddo
allocate(array(ao_num, ao_num, ao_num, ao_num))
array = 0.d0
allocate(array_ref(ao_num, ao_num, ao_num, ao_num))
@ -336,6 +301,9 @@ subroutine routine_v_ij_u_cst_mu_j1b_test
enddo
enddo
enddo
print*,'******'
print*,'******'
print*,'routine_v_ij_u_cst_mu_j1b_test'
print*,'accu_abs = ',accu_abs/dble(ao_num)**4
print*,'accu_relat = ',accu_relat/dble(ao_num)**4
@ -427,19 +395,6 @@ subroutine routine_int2_u2_j1b2
integer :: i,j,ipoint,k,l
double precision :: weight,accu_relat, accu_abs, contrib
double precision, allocatable :: array(:,:,:,:), array_ref(:,:,:,:)
! print*,'ao_overlap_abs = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_overlap_abs(i,:)
! enddo
! print*,'center = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_prod_center(2,i,:)
! enddo
! print*,'sigma = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_prod_sigma(i,:)
! enddo
allocate(array(ao_num, ao_num, ao_num, ao_num))
array = 0.d0
@ -473,6 +428,9 @@ subroutine routine_int2_u2_j1b2
enddo
enddo
enddo
print*,'******'
print*,'******'
print*,'routine_int2_u2_j1b2'
print*,'accu_abs = ',accu_abs/dble(ao_num)**4
print*,'accu_relat = ',accu_relat/dble(ao_num)**4
@ -486,19 +444,6 @@ subroutine routine_int2_u_grad1u_x_j1b2
integer :: i,j,ipoint,k,l,m
double precision :: weight,accu_relat, accu_abs, contrib
double precision, allocatable :: array(:,:,:,:), array_ref(:,:,:,:)
! print*,'ao_overlap_abs = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_overlap_abs(i,:)
! enddo
! print*,'center = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_prod_center(2,i,:)
! enddo
! print*,'sigma = '
! do i = 1, ao_num
! write(*,'(100(F10.5,X))')ao_prod_sigma(i,:)
! enddo
allocate(array(ao_num, ao_num, ao_num, ao_num))
array = 0.d0
@ -534,6 +479,9 @@ subroutine routine_int2_u_grad1u_x_j1b2
enddo
enddo
enddo
print*,'******'
print*,'******'
print*,'routine_int2_u_grad1u_x_j1b2'
print*,'accu_abs = ',accu_abs/dble(ao_num)**4
print*,'accu_relat = ',accu_relat/dble(ao_num)**4
@ -579,6 +527,9 @@ subroutine routine_v_ij_u_cst_mu_j1b
enddo
enddo
enddo
print*,'******'
print*,'******'
print*,'routine_v_ij_u_cst_mu_j1b'
print*,'accu_abs = ',accu_abs/dble(ao_num)**4
print*,'accu_relat = ',accu_relat/dble(ao_num)**4
@ -742,6 +693,9 @@ subroutine test_total_grad_lapl
enddo
enddo
enddo
print*,'******'
print*,'******'
print*,' test_total_grad_lapl'
print*,'accu_abs = ',accu_abs/dble(ao_num)**4
print*,'accu_relat = ',accu_relat/dble(ao_num)**4
@ -767,6 +721,9 @@ subroutine test_total_grad_square
enddo
enddo
enddo
print*,'******'
print*,'******'
print*,'test_total_grad_square'
print*,'accu_abs = ',accu_abs/dble(ao_num)**4
print*,'accu_relat = ',accu_relat/dble(ao_num)**4
@ -1057,3 +1014,86 @@ end
! ---
subroutine test_old_ints
implicit none
integer :: i,j,k,l
double precision :: old, new, contrib, get_ao_tc_sym_two_e_pot
double precision :: integral_sym , integral_nsym,accu
PROVIDE ao_tc_sym_two_e_pot_in_map
accu = 0.d0
do j = 1, ao_num
do l= 1, ao_num
do i = 1, ao_num
do k = 1, ao_num
! integral_sym = get_ao_tc_sym_two_e_pot(i, j, k, l, ao_tc_sym_two_e_pot_map)
! ao_non_hermit_term_chemist(k,i,l,j) = < k l | [erf( mu r12) - 1] d/d_r12 | i j > on the AO basis
! integral_nsym = ao_non_hermit_term_chemist(k,i,l,j)
! old = integral_sym + integral_nsym
! old = tc_grad_square_ao(k,i,l,j) + tc_grad_and_lapl_ao(k,i,l,j) + ao_two_e_coul(k,i,l,j)
new = ao_tc_int_chemist_test(k,i,l,j)
old = ao_tc_int_chemist_no_cycle(k,i,l,j)
contrib = dabs(old - new)
if(contrib.gt.1.d-6)then
print*,'problem !!'
print*,i,j,k,l
print*,old, new, contrib
endif
accu += contrib
enddo
enddo
enddo
enddo
print*,'******'
print*,'******'
print*,'in test_old_ints'
print*,'accu = ',accu/dble(ao_num**4)
end
subroutine test_int2_grad1_u12_ao_test
implicit none
integer :: i,j,ipoint,m,k,l
double precision :: weight,accu_relat, accu_abs, contrib
double precision, allocatable :: array(:,:,:,:), array_ref(:,:,:,:)
allocate(array(ao_num, ao_num, ao_num, ao_num))
array = 0.d0
allocate(array_ref(ao_num, ao_num, ao_num, ao_num))
array_ref = 0.d0
do m = 1, 3
do ipoint = 1, n_points_final_grid
weight = final_weight_at_r_vector(ipoint)
do k = 1, ao_num
do l = 1, ao_num
do i = 1, ao_num
do j = 1, ao_num
array(j,i,l,k) += int2_grad1_u12_ao_test(j,i,ipoint,m) * aos_grad_in_r_array_transp(m,k,ipoint) * aos_in_r_array(l,ipoint) * weight
array_ref(j,i,l,k) += int2_grad1_u12_ao(j,i,ipoint,m) * aos_grad_in_r_array_transp(m,k,ipoint) * aos_in_r_array(l,ipoint) * weight
enddo
enddo
enddo
enddo
enddo
enddo
accu_relat = 0.d0
accu_abs = 0.d0
do k = 1, ao_num
do l = 1, ao_num
do i = 1, ao_num
do j = 1, ao_num
contrib = dabs(array(j,i,l,k) - array_ref(j,i,l,k))
accu_abs += contrib
if(dabs(array_ref(j,i,l,k)).gt.1.d-10)then
accu_relat += contrib/dabs(array_ref(j,i,l,k))
endif
enddo
enddo
enddo
enddo
print*,'******'
print*,'******'
print*,'test_int2_grad1_u12_ao_test'
print*,'accu_abs = ',accu_abs/dble(ao_num)**4
print*,'accu_relat = ',accu_relat/dble(ao_num)**4
end

20
src/tools/sort_wf.irp.f Normal file
View File

@ -0,0 +1,20 @@
program sort_wf
implicit none
read_wf = .true.
touch read_wf
call routine
end
subroutine routine
implicit none
integer :: i
character*(128) :: output
integer :: i_unit_output,getUnitAndOpen
output=trim(ezfio_filename)//'.wf_sorted'
i_unit_output = getUnitAndOpen(output,'w')
do i = 1, N_det
write(i_unit_output, *)i,dabs(psi_coef_sorted(i,1))/dabs(psi_coef_sorted(1,1))
enddo
end