program delta_3bSVD_v1

  BEGIN_DOC
  ! !!!
  END_DOC

  implicit none

  read_wf = .True.
  touch read_wf

  call run()

end 



subroutine run()

  implicit none

  integer                       :: ii, ii_i, ii_j, k, l
  integer                       :: na, nb, nsvd
  integer                       :: n_tmp, n_TSVD
  double precision, allocatable :: U(:,:), V(:,:), S(:)

  double precision, allocatable :: delta0_det(:), delta1_det(:)
  double precision, allocatable :: deltaSVD1_kl(:,:), deltaSVD2_kl(:,:)
  double precision, allocatable :: delta_tmp(:)

  na = n_det_alpha_unique
  nb = n_det_beta_unique

  call ezfio_get_spindeterminants_n_svd_coefs(nsvd)

  ! read SVD vectors
  allocate( u(na,nsvd) , v(nb,nsvd) )
  call ezfio_get_spindeterminants_psi_svd_alpha(U)
  call ezfio_get_spindeterminants_psi_svd_beta (V)

!  allocate( s(nsvd) )
!  call ezfio_get_spindeterminants_psi_svd_coefs(S)

  allocate( delta0_det(N_det) )
  call ezfio_get_dmc_dress_dmc_delta_h(delta0_det)


  ! ---------------------------------------------
  !               parameters

  n_TSVD = 50

  print *, " ---  delta_3bSVD_v1  ---"
  print *, " n_TSVD =", n_TSVD
  print *, " tot    =", nsvd * n_TSVD + &
                        (nsvd-n_TSVD)*n_TSVD 

  !
  ! ---------------------------------------------

  n_tmp = n_TSVD + 1

  allocate( deltaSVD1_kl(1:nsvd,1:n_TSVD) , deltaSVD2_kl(1:n_TSVD,n_tmp:nsvd) )
  call obtenir_deltaSVD_3b(nsvd, n_TSVD, U, V, delta0_det, deltaSVD1_kl, deltaSVD2_kl)
  deallocate( delta0_det )

  ! [ delta ]_mn = \sum_{k,l}  U_{m,k} delta_SVD_{k,l} V_{n,l}
  allocate( delta1_det(N_det) )
  delta1_det(1:N_det) = 0.d0

 !$OMP PARALLEL DEFAULT(NONE)                                                                &
 !$OMP          SHARED(nsvd, n_TSVD, n_tmp, N_det, U, V, delta1_det, deltaSVD1_kl            &
 !$OMP                , deltaSVD2_kl, psi_bilinear_matrix_rows, psi_bilinear_matrix_columns) &
 !$OMP          PRIVATE(k, l, ii, ii_i, ii_j, delta_tmp)
  allocate( delta_tmp(N_det) )
  delta_tmp(1:N_det) = 0.d0
 !$OMP DO
  do ii = 1, N_det
    ii_i = psi_bilinear_matrix_rows   (ii)
    ii_j = psi_bilinear_matrix_columns(ii)
    do l = 1, n_TSVD
      do k = 1, nsvd
        delta_tmp(ii) = delta_tmp(ii) + U(ii_i,k) * V(ii_j,l) * deltaSVD1_kl(k,l)
      enddo
    enddo
    do l = n_tmp, nsvd
      do k = 1, n_TSVD
        delta_tmp(ii) = delta_tmp(ii) + U(ii_i,k) * V(ii_j,l) * deltaSVD2_kl(k,l)
      enddo
    enddo
  enddo
 !$OMP END DO
 !$OMP CRITICAL
  do ii = 1, N_det
    delta1_det(ii) = delta1_det(ii) + delta_tmp(ii)
  enddo
 !$OMP END CRITICAL
  deallocate( delta_tmp )
 !$OMP END PARALLEL

!  deallocate( S )
  deallocate( U , V )
  deallocate( deltaSVD1_kl, deltaSVD2_kl )

  call ezfio_set_dmc_dress_dmc_delta_h(delta1_det)
  deallocate( delta1_det )

end





! _______________________________________________________________________________________________________
!
!     [ delta_SVD ]_kl = < dup_SVD ddn_SVD | H_tilde - H | psi >
!                      = \sum_{i,j}  U_{i,k} V_{j,l} < dup_i ddn_j | H_tilde - H | psi >
!
subroutine obtenir_deltaSVD_3b(nsvd, n_TSVD, U, V, delta0_det, deltaSVD1_kl, deltaSVD2_kl)

  implicit none

  integer,          intent(in)  :: nsvd, n_TSVD
  double precision, intent(in)  :: U(n_det_alpha_unique,nsvd), V(n_det_beta_unique,nsvd)
  double precision, intent(in)  :: delta0_det(N_det)

  double precision, intent(out) :: deltaSVD1_kl(1:nsvd,1:n_TSVD)
  double precision, intent(out) :: deltaSVD2_kl(1:n_TSVD,n_TSVD+1:nsvd)

  integer                       :: n_tmp
  integer                       :: k, l, ii, ii_i, ii_j
  double precision, allocatable :: delta_tmp(:,:)

  n_tmp = n_TSVD + 1
  deltaSVD1_kl(1:nsvd,1:n_TSVD)     = 0.d0
  deltaSVD2_kl(1:n_TSVD,n_tmp:nsvd) = 0.d0

 !$OMP PARALLEL DEFAULT(NONE)                                                  &
 !$OMP          SHARED(nsvd, n_TSVD, N_det, U, V, delta0_det, deltaSVD1_kl     &
 !$OMP                , psi_bilinear_matrix_rows, psi_bilinear_matrix_columns) &
 !$OMP          PRIVATE(k, l, ii, ii_i, ii_j, delta_tmp)
  allocate( delta_tmp(1:nsvd,1:n_TSVD) )
  delta_tmp(1:nsvd,1:n_TSVD) = 0.d0
 !$OMP DO
  do l = 1, n_TSVD
    do k = 1, nsvd
      do ii = 1, N_det
        ii_i = psi_bilinear_matrix_rows   (ii)
        ii_j = psi_bilinear_matrix_columns(ii)
        delta_tmp(k,l) = delta_tmp(k,l) + U(ii_i,k) * V(ii_j,l) * delta0_det(ii)
      enddo
    enddo
  enddo
 !$OMP END DO
 !$OMP CRITICAL
  do l = 1, n_TSVD
    do k = 1, nsvd
      deltaSVD1_kl(k,l) = deltaSVD1_kl(k,l) + delta_tmp(k,l)
    enddo
  enddo
 !$OMP END CRITICAL
  deallocate(delta_tmp)
 !$OMP END PARALLEL

 !$OMP PARALLEL DEFAULT(NONE)                                                     &
 !$OMP          SHARED(nsvd, n_TSVD, n_tmp, N_det, U, V, delta0_det, deltaSVD2_kl &
 !$OMP                , psi_bilinear_matrix_rows, psi_bilinear_matrix_columns)    &
 !$OMP          PRIVATE(k, l, ii, ii_i, ii_j, delta_tmp)
  allocate( delta_tmp(1:n_TSVD,n_tmp:nsvd) )
  delta_tmp(1:n_TSVD,n_tmp:nsvd) = 0.d0
 !$OMP DO
  do l = n_tmp, nsvd
    do k = 1, n_TSVD
      do ii = 1, N_det
        ii_i = psi_bilinear_matrix_rows   (ii)
        ii_j = psi_bilinear_matrix_columns(ii)
        delta_tmp(k,l) = delta_tmp(k,l) + U(ii_i,k) * V(ii_j,l) * delta0_det(ii)
      enddo
    enddo
  enddo
 !$OMP END DO
 !$OMP CRITICAL
  do l = n_tmp, nsvd
    do k = 1, n_TSVD
      deltaSVD2_kl(k,l) = deltaSVD2_kl(k,l) + delta_tmp(k,l)
    enddo
  enddo
 !$OMP END CRITICAL
  deallocate(delta_tmp)
 !$OMP END PARALLEL

  return
end subroutine obtenir_deltaSVD_3b
! _______________________________________________________________________________________________________
! _______________________________________________________________________________________________________