davidson¶
Abstract module for Davidson’s diagonalization. It contains everything required for the Davidson algorithm, dressed or not. If a dressing is used, the dressing column should be defined and the davidson_dressed module should be used. If no dressing is required, the davidson module should be used, and it has a default zero dressing vector.
The important providers for that module are:
# psi_energy which is the expectation value over the wave function (psi_det, psi_coef) of the Hamiltonian, dressed or not. It uses the general subroutine u_0_H_u_0.
# psi_energy_two_e which is the expectation value over the wave function (psi_det, psi_coef) of the standard two-electrons coulomb operator. It uses the general routine u_0_H_u_0_two_e.
EZFIO parameters¶
-
threshold_davidson¶ Thresholds of Davidson’s algorithm
Default: 1.e-10
-
n_states_diag¶ Number of states to consider during the Davdison diagonalization
Default: 4
-
davidson_sze_max¶ Number of micro-iterations before re-contracting
Default: 8
-
state_following¶ If
true, the states are re-ordered to match the input statesDefault: False
-
disk_based_davidson¶ If
true, a memory-mapped file may be used to store the W and S2 vectors if not enough RAM is availableDefault: True
-
distributed_davidson¶ If
true, use the distributed algorithmDefault: True
-
only_expected_s2¶ If
true, use filter out all vectors with bad \(\widehat{S^2}\) valuesDefault: True
Providers¶
-
ci_eigenvectors¶ File :
davidson/diagonalize_ci.irp.fdouble precision, allocatable :: ci_electronic_energy (N_states_diag) double precision, allocatable :: ci_eigenvectors (N_det,N_states_diag) double precision, allocatable :: ci_s2 (N_states_diag)
Eigenvectors/values of the CI matrix
Needs:
n_intn_statesn_states_diagnthreads_davidsonpsi_coef
psi_dets2_eigs2_matrix_all_detss_zthreshold_davidson
Needed by:
-
ci_electronic_energy¶ File :
davidson/diagonalize_ci.irp.fdouble precision, allocatable :: ci_electronic_energy (N_states_diag) double precision, allocatable :: ci_eigenvectors (N_det,N_states_diag) double precision, allocatable :: ci_s2 (N_states_diag)
Eigenvectors/values of the CI matrix
Needs:
n_intn_statesn_states_diagnthreads_davidsonpsi_coef
psi_dets2_eigs2_matrix_all_detss_zthreshold_davidson
Needed by:
-
ci_energy¶ File :
davidson/diagonalize_ci.irp.fdouble precision, allocatable :: ci_energy (N_states_diag)
n_stateslowest eigenvalues of the CI matrixNeeds:
n_statesn_states_diag
-
ci_s2¶ File :
davidson/diagonalize_ci.irp.fdouble precision, allocatable :: ci_electronic_energy (N_states_diag) double precision, allocatable :: ci_eigenvectors (N_det,N_states_diag) double precision, allocatable :: ci_s2 (N_states_diag)
Eigenvectors/values of the CI matrix
Needs:
n_intn_statesn_states_diagnthreads_davidsonpsi_coef
psi_dets2_eigs2_matrix_all_detss_zthreshold_davidson
Needed by:
-
davidson_criterion¶ File :
davidson/parameters.irp.fcharacter(64) :: davidson_criterion
Can be : [ energy | residual | both | wall_time | cpu_time | iterations ]
-
dressed_column_idx¶ File :
davidson/diagonalization_hs2_dressed.irp.finteger, allocatable :: dressed_column_idx (N_states)
Index of the dressed columns
Needs:
n_states
-
n_states_diag¶ File :
davidson/input.irp.finteger :: n_states_diag
Number of states to consider during the Davdison diagonalization
Needs:
n_states
Needed by:
-
nthreads_davidson¶ File :
davidson/davidson_parallel.irp.finteger :: nthreads_davidson
Number of threads for Davidson
Needs:
Needed by:
-
psi_energy¶ File :
davidson/u0_h_u0.irp.fdouble precision, allocatable :: psi_energy (N_states) double precision, allocatable :: psi_s2 (N_states)
psi_energy(i) = \(\langle \Psi_i | H | \Psi_i \rangle\)
psi_s2(i) = \(\langle \Psi_i | S^2 | \Psi_i \rangle\)
Needs:
n_statesn_states_diagpsi_coef
Needed by:
-
psi_energy_two_e¶ File :
davidson/u0_wee_u0.irp.fdouble precision, allocatable :: psi_energy_two_e (N_states)
Energy of the current wave function
Needs:
-
psi_energy_with_nucl_rep¶ File :
davidson/u0_h_u0.irp.fdouble precision, allocatable :: psi_energy_with_nucl_rep (N_states)
Energy of the wave function with the nuclear repulsion energy.
Needs:
n_states
-
psi_s2¶ File :
davidson/u0_h_u0.irp.fdouble precision, allocatable :: psi_energy (N_states) double precision, allocatable :: psi_s2 (N_states)
psi_energy(i) = \(\langle \Psi_i | H | \Psi_i \rangle\)
psi_s2(i) = \(\langle \Psi_i | S^2 | \Psi_i \rangle\)
Needs:
n_statesn_states_diagpsi_coef
Needed by:
Subroutines / functions¶
-
davidson_collector:()¶ File :
davidson/davidson_parallel.irp.fsubroutine davidson_collector(zmq_to_qp_run_socket, zmq_socket_pull, v0, s0, sze, N_st)
Routine collecting the results of the workers in Davidson’s algorithm.
Needs:
Called by:
h_s2_u_0_nstates_zmq()
Calls:
davidson_pull_results()
-
davidson_converged:()¶ File :
davidson/parameters.irp.fsubroutine davidson_converged(energy,residual,wall,iterations,cpu,N_st,converged)
True if the Davidson algorithm is converged
Needs:
threshold_davidson
Called by:
davidson_diag_hjj_sjj()
Calls:
cpu_time()
wall_time()
-
davidson_diag_hjj_sjj:()¶ File :
davidson/diagonalization_hs2_dressed.irp.fsubroutine davidson_diag_hjj_sjj(dets_in,u_in,H_jj,s2_out,energies,dim_in,sze,N_st,N_st_diag_in,Nint,dressing_state,converged)
Davidson diagonalization with specific diagonal elements of the H matrix
H_jj : specific diagonal H matrix elements to diagonalize de Davidson
S2_out : Output : s^2
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
Needs:
dressing_column_hezfio_work_dirdavidson_sze_maxstate_followingpsi_det_alpha_uniquenuclear_repulsionnprocqp_max_mem
disk_based_davidsons2_eigpsi_det_beta_uniqueonly_expected_s2distributed_davidsonn_statesn_int
Called by:
davidson_diag_hs2()
Calls:
c_f_pointer()check_mem()cpu_time()davidson_converged()dgemm()dswap()h_s2_u_0_nstates_openmp()
h_s2_u_0_nstates_zmq()lapack_diag()mmap()munmap()normalize()ortho_qr()random_number()
resident_memory()sgemm()wall_time()write_double()write_int()write_time()
Touches:
-
davidson_diag_hs2:()¶ File :
davidson/diagonalization_hs2_dressed.irp.fsubroutine davidson_diag_hs2(dets_in,u_in,s2_out,dim_in,energies,sze,N_st,N_st_diag,Nint,dressing_state,converged)
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
Needs:
Called by:
Calls:
davidson_diag_hjj_sjj()
Touches:
-
davidson_pull_results:()¶ File :
davidson/davidson_parallel.irp.fsubroutine davidson_pull_results(zmq_socket_pull, v_t, s_t, imin, imax, task_id)
Pull the results of $H|U rangle$ on the master.
Needs:
Called by:
davidson_collector()
-
davidson_push_results:()¶ File :
davidson/davidson_parallel.irp.fsubroutine davidson_push_results(zmq_socket_push, v_t, s_t, imin, imax, task_id)
Push the results of $H|U rangle$ from a worker to the master.
Needs:
Called by:
davidson_slave_work()
-
davidson_run_slave:()¶ File :
davidson/davidson_parallel.irp.fsubroutine davidson_run_slave(thread,iproc)
Slave routine for Davidson’s diagonalization.
Needs:
Called by:
davidson_slave_inproc()
davidson_slave_tcp()
Calls:
davidson_slave_work()end_zmq_push_socket()
end_zmq_to_qp_run_socket()
sleep()
-
davidson_slave_inproc:()¶ File :
davidson/davidson_parallel.irp.fsubroutine davidson_slave_inproc(i)
Called by:
h_s2_u_0_nstates_zmq()
Calls:
davidson_run_slave()
-
davidson_slave_tcp:()¶ File :
davidson/davidson_parallel.irp.fsubroutine davidson_slave_tcp(i)
Called by:
run_slave_main()
Calls:
davidson_run_slave()
-
davidson_slave_work:()¶ File :
davidson/davidson_parallel.irp.fsubroutine davidson_slave_work(zmq_to_qp_run_socket, zmq_socket_push, N_st, sze, worker_id)
Needs:
Called by:
davidson_run_slave()
Calls:
davidson_push_results()
h_s2_u_0_nstates_openmp_work()
sleep()
-
diagonalize_ci:()¶ File :
davidson/diagonalize_ci.irp.fsubroutine diagonalize_CI
Replace the coefficients of the CI states by the coefficients of the eigenstates of the CI matrix.
Needs:
psi_coefci_electronic_energyn_states
Called by:
remove_small_contributions()
run_cipsi()
run_stochastic_cipsi()
Touches:
-
h_s2_u_0_nstates_openmp:()¶ File :
davidson/u0_h_u0.irp.fsubroutine H_S2_u_0_nstates_openmp(v_0,s_0,u_0,N_st,sze)
Computes $v_0 = H|u_0rangle$ and $s_0 = S^2 |u_0rangle$.
Assumes that the determinants are in psi_det
istart, iend, ishift, istep are used in ZMQ parallelization.
Needs:
Called by:
davidson_diag_hjj_sjj()
u_0_h_u_0()
Calls:
dset_order()
dtranspose()
h_s2_u_0_nstates_openmp_work()
-
h_s2_u_0_nstates_openmp_work:()¶ File :
davidson/u0_h_u0.irp.fsubroutine H_S2_u_0_nstates_openmp_work(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep)
Computes $v_t = H|u_trangle$ and $s_t = S^2 |u_trangle$
Default should be 1,N_det,0,1
Needs:
Called by:
davidson_slave_work()
h_s2_u_0_nstates_openmp()
Calls:
h_s2_u_0_nstates_openmp_work_1()h_s2_u_0_nstates_openmp_work_2()
h_s2_u_0_nstates_openmp_work_3()h_s2_u_0_nstates_openmp_work_4()
h_s2_u_0_nstates_openmp_work_n_int()
-
h_s2_u_0_nstates_openmp_work_1:()¶ File :
davidson/u0_h_u0.irp.f_template_631subroutine H_S2_u_0_nstates_openmp_work_1(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep)
Computes $v_t = H|u_t
angle$ and $s_t = S^2 |u_t angle$
Default should be 1,N_det,0,1
Needs:
Called by:
h_s2_u_0_nstates_openmp_work()
Calls:
get_all_spin_singles_1()get_all_spin_singles_and_doubles_1()
get_s2()i_h_j_double_alpha_beta()
i_h_j_double_spin()i_h_j_mono_spin()
-
h_s2_u_0_nstates_openmp_work_2:()¶ File :
davidson/u0_h_u0.irp.f_template_631subroutine H_S2_u_0_nstates_openmp_work_2(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep)
Computes $v_t = H|u_t
angle$ and $s_t = S^2 |u_t angle$
Default should be 1,N_det,0,1
Needs:
Called by:
h_s2_u_0_nstates_openmp_work()
Calls:
get_all_spin_singles_2()get_all_spin_singles_and_doubles_2()
get_s2()i_h_j_double_alpha_beta()
i_h_j_double_spin()i_h_j_mono_spin()
-
h_s2_u_0_nstates_openmp_work_3:()¶ File :
davidson/u0_h_u0.irp.f_template_631subroutine H_S2_u_0_nstates_openmp_work_3(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep)
Computes $v_t = H|u_t
angle$ and $s_t = S^2 |u_t angle$
Default should be 1,N_det,0,1
Needs:
Called by:
h_s2_u_0_nstates_openmp_work()
Calls:
get_all_spin_singles_3()get_all_spin_singles_and_doubles_3()
get_s2()i_h_j_double_alpha_beta()
i_h_j_double_spin()i_h_j_mono_spin()
-
h_s2_u_0_nstates_openmp_work_4:()¶ File :
davidson/u0_h_u0.irp.f_template_631subroutine H_S2_u_0_nstates_openmp_work_4(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep)
Computes $v_t = H|u_t
angle$ and $s_t = S^2 |u_t angle$
Default should be 1,N_det,0,1
Needs:
Called by:
h_s2_u_0_nstates_openmp_work()
Calls:
get_all_spin_singles_4()get_all_spin_singles_and_doubles_4()
get_s2()i_h_j_double_alpha_beta()
i_h_j_double_spin()i_h_j_mono_spin()
-
h_s2_u_0_nstates_openmp_work_n_int:()¶ File :
davidson/u0_h_u0.irp.f_template_631subroutine H_S2_u_0_nstates_openmp_work_N_int(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep)
Computes $v_t = H|u_t
angle$ and $s_t = S^2 |u_t angle$
Default should be 1,N_det,0,1
Needs:
Called by:
h_s2_u_0_nstates_openmp_work()
Calls:
get_all_spin_singles_and_doubles_n_int()get_all_spin_singles_n_int()
get_s2()i_h_j_double_alpha_beta()
i_h_j_double_spin()i_h_j_mono_spin()
-
h_s2_u_0_nstates_zmq:()¶ File :
davidson/davidson_parallel.irp.fsubroutine H_S2_u_0_nstates_zmq(v_0,s_0,u_0,N_st,sze)
Computes $v_0 = H|u_0rangle$ and $s_0 = S^2 |u_0rangle$
n : number of determinants
H_jj : array of $langle j|H|j rangle$
S2_jj : array of $langle j|S^2|j rangle$
Needs:
Called by:
davidson_diag_hjj_sjj()
u_0_h_u_0()
Calls:
davidson_collector()davidson_slave_inproc()dset_order()
dtranspose()end_parallel_job()
new_parallel_job()omp_set_nested()
Touches:
-
h_s2_u_0_two_e_nstates_openmp:()¶ File :
davidson/u0_wee_u0.irp.fsubroutine H_S2_u_0_two_e_nstates_openmp(v_0,s_0,u_0,N_st,sze)
Computes $v_0 = H|u_0rangle$ and $s_0 = S^2 |u_0rangle$
Assumes that the determinants are in psi_det
istart, iend, ishift, istep are used in ZMQ parallelization.
Needs:
Called by:
u_0_h_u_0_two_e()
Calls:
dset_order()
dtranspose()
h_s2_u_0_two_e_nstates_openmp_work()
-
h_s2_u_0_two_e_nstates_openmp_work:()¶ File :
davidson/u0_wee_u0.irp.fsubroutine H_S2_u_0_two_e_nstates_openmp_work(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep)
Computes $v_t = H|u_trangle$ and $s_t = S^2 |u_trangle$
Default should be 1,N_det,0,1
Needs:
Called by:
h_s2_u_0_two_e_nstates_openmp()
Calls:
h_s2_u_0_two_e_nstates_openmp_work_1()h_s2_u_0_two_e_nstates_openmp_work_2()
h_s2_u_0_two_e_nstates_openmp_work_3()h_s2_u_0_two_e_nstates_openmp_work_4()
h_s2_u_0_two_e_nstates_openmp_work_n_int()
-
h_s2_u_0_two_e_nstates_openmp_work_1:()¶ File :
davidson/u0_wee_u0.irp.f_template_457subroutine H_S2_u_0_two_e_nstates_openmp_work_1(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep)
Computes $v_t = H|u_t
angle$ and $s_t = S^2 |u_t angle$
Default should be 1,N_det,0,1
Needs:
Called by:
h_s2_u_0_two_e_nstates_openmp_work()
Calls:
get_all_spin_singles_1()get_all_spin_singles_and_doubles_1()
get_s2()i_h_j_double_alpha_beta()
i_h_j_double_spin()i_wee_j_mono()
-
h_s2_u_0_two_e_nstates_openmp_work_2:()¶ File :
davidson/u0_wee_u0.irp.f_template_457subroutine H_S2_u_0_two_e_nstates_openmp_work_2(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep)
Computes $v_t = H|u_t
angle$ and $s_t = S^2 |u_t angle$
Default should be 1,N_det,0,1
Needs:
Called by:
h_s2_u_0_two_e_nstates_openmp_work()
Calls:
get_all_spin_singles_2()get_all_spin_singles_and_doubles_2()
get_s2()i_h_j_double_alpha_beta()
i_h_j_double_spin()i_wee_j_mono()
-
h_s2_u_0_two_e_nstates_openmp_work_3:()¶ File :
davidson/u0_wee_u0.irp.f_template_457subroutine H_S2_u_0_two_e_nstates_openmp_work_3(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep)
Computes $v_t = H|u_t
angle$ and $s_t = S^2 |u_t angle$
Default should be 1,N_det,0,1
Needs:
Called by:
h_s2_u_0_two_e_nstates_openmp_work()
Calls:
get_all_spin_singles_3()get_all_spin_singles_and_doubles_3()
get_s2()i_h_j_double_alpha_beta()
i_h_j_double_spin()i_wee_j_mono()
-
h_s2_u_0_two_e_nstates_openmp_work_4:()¶ File :
davidson/u0_wee_u0.irp.f_template_457subroutine H_S2_u_0_two_e_nstates_openmp_work_4(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep)
Computes $v_t = H|u_t
angle$ and $s_t = S^2 |u_t angle$
Default should be 1,N_det,0,1
Needs:
Called by:
h_s2_u_0_two_e_nstates_openmp_work()
Calls:
get_all_spin_singles_4()get_all_spin_singles_and_doubles_4()
get_s2()i_h_j_double_alpha_beta()
i_h_j_double_spin()i_wee_j_mono()
-
h_s2_u_0_two_e_nstates_openmp_work_n_int:()¶ File :
davidson/u0_wee_u0.irp.f_template_457subroutine H_S2_u_0_two_e_nstates_openmp_work_N_int(v_t,s_t,u_t,N_st,sze,istart,iend,ishift,istep)
Computes $v_t = H|u_t
angle$ and $s_t = S^2 |u_t angle$
Default should be 1,N_det,0,1
Needs:
Called by:
h_s2_u_0_two_e_nstates_openmp_work()
Calls:
get_all_spin_singles_and_doubles_n_int()get_all_spin_singles_n_int()
get_s2()i_h_j_double_alpha_beta()
i_h_j_double_spin()i_wee_j_mono()
-
u_0_h_u_0:()¶ File :
davidson/u0_h_u0.irp.fsubroutine u_0_H_u_0(e_0,s_0,u_0,n,keys_tmp,Nint,N_st,sze)
Computes $E_0 = frac{langle u_0|H|u_0 rangle}{langle u_0|u_0 rangle}$
and $S_0 = frac{langle u_0|S^2|u_0 rangle}{langle u_0|u_0 rangle}$
n : number of determinants
Needs:
n_states
distributed_davidson
Called by:
Calls:
h_s2_u_0_nstates_openmp()
h_s2_u_0_nstates_zmq()
Touches:
-
u_0_h_u_0_two_e:()¶ File :
davidson/u0_wee_u0.irp.fsubroutine u_0_H_u_0_two_e(e_0,u_0,n,keys_tmp,Nint,N_st,sze)
Computes $E_0 = frac{ langle u_0|H|u_0rangle}{langle u_0|u_0 rangle}$.
n : number of determinants
Called by:
Calls:
h_s2_u_0_two_e_nstates_openmp()
-
zmq_get_n_states_diag:()¶ File :
davidson/davidson_parallel.irp.finteger function zmq_get_N_states_diag(zmq_to_qp_run_socket, worker_id)
Get N_states_diag from the qp_run scheduler
Needs:
Touches:
-
zmq_put_n_states_diag:()¶ File :
davidson/davidson_parallel.irp.finteger function zmq_put_N_states_diag(zmq_to_qp_run_socket,worker_id)
Put N_states_diag on the qp_run scheduler
Needs: