From e710d2623765dce23896f576cb0b1a37b393a9f1 Mon Sep 17 00:00:00 2001 From: Anthony Scemama Date: Fri, 17 Mar 2023 19:23:07 +0100 Subject: [PATCH] Improve I/O on TC integrals --- src/non_h_ints_mu/grad_squared.irp.f | 70 ++++++--------- src/non_h_ints_mu/grad_squared_manu.irp.f | 76 +++++++--------- src/non_h_ints_mu/new_grad_tc.irp.f | 28 ++---- src/non_h_ints_mu/new_grad_tc_manu.irp.f | 103 ++++++++-------------- src/tc_keywords/EZFIO.cfg | 34 +++---- 5 files changed, 118 insertions(+), 193 deletions(-) diff --git a/src/non_h_ints_mu/grad_squared.irp.f b/src/non_h_ints_mu/grad_squared.irp.f index 7925fa7c..1fd39f6a 100644 --- a/src/non_h_ints_mu/grad_squared.irp.f +++ b/src/non_h_ints_mu/grad_squared.irp.f @@ -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 !!$OMP END PARALLEL endif - if(write_tc_integ) then - open(unit=11, form="unformatted", file='tc_grad_square_ao', action="write") - do i = 1, ao_num - do j = 1, ao_num - do k = 1, ao_num - do l = 1, ao_num - write(11) tc_grad_square_ao(l,k,j,i) - enddo - enddo - enddo - enddo + if(write_tc_integ.and.mpi_master) then + open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/tc_grad_square_ao', action="write") + call ezfio_set_work_empty(.False.) + write(11) tc_grad_square_ao close(11) + call ezfio_set_tc_keywords_io_tc_integ('Read') endif call wall_time(time1) print*, ' Wall time for tc_grad_square_ao = ', time1 - time0 -END_PROVIDER +END_PROVIDER ! --- diff --git a/src/non_h_ints_mu/grad_squared_manu.irp.f b/src/non_h_ints_mu/grad_squared_manu.irp.f index cb9e15c4..66f3c693 100644 --- a/src/non_h_ints_mu/grad_squared_manu.irp.f +++ b/src/non_h_ints_mu/grad_squared_manu.irp.f @@ -17,29 +17,21 @@ BEGIN_PROVIDER [double precision, tc_grad_square_ao_test, (ao_num, ao_num, ao_nu call wall_time(time0) if(read_tc_integ) then - - open(unit=11, form="unformatted", file='tc_grad_square_ao_test', action="read") - do i = 1, ao_num - do j = 1, ao_num - do k = 1, ao_num - do l = 1, ao_num - read(11) tc_grad_square_ao_test(l,k,j,i) - enddo - enddo - enddo - enddo + + open(unit=11, form="unformatted", file=trim(ezfio_filename)//'/work/tc_grad_square_ao_test', action="read") + read(11) tc_grad_square_ao_test close(11) else provide u12sq_j1bsq_test u12_grad1_u12_j1b_grad1_j1b_test grad12_j12_test - + allocate(b_mat(n_points_final_grid,ao_num,ao_num), tmp(ao_num,ao_num,n_points_final_grid)) - + b_mat = 0.d0 !$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 ! --- diff --git a/src/non_h_ints_mu/new_grad_tc.irp.f b/src/non_h_ints_mu/new_grad_tc.irp.f index a15f690a..754e1240 100644 --- a/src/non_h_ints_mu/new_grad_tc.irp.f +++ b/src/non_h_ints_mu/new_grad_tc.irp.f @@ -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) diff --git a/src/non_h_ints_mu/new_grad_tc_manu.irp.f b/src/non_h_ints_mu/new_grad_tc_manu.irp.f index 47b05e52..901e3048 100644 --- a/src/non_h_ints_mu/new_grad_tc_manu.irp.f +++ b/src/non_h_ints_mu/new_grad_tc_manu.irp.f @@ -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 ! --- diff --git a/src/tc_keywords/EZFIO.cfg b/src/tc_keywords/EZFIO.cfg index e397e700..3f34c088 100644 --- a/src/tc_keywords/EZFIO.cfg +++ b/src/tc_keywords/EZFIO.cfg @@ -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 @@ -60,11 +60,11 @@ 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 +78,7 @@ 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 +126,7 @@ 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 +152,7 @@ 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 +180,15 @@ 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