10
0
mirror of https://github.com/QuantumPackage/qp2.git synced 2024-11-14 18:13:51 +01:00

added cipsi_tc_bi_ortho

This commit is contained in:
eginer 2022-10-24 21:14:31 +02:00
parent db02a9c865
commit 7d509dff61
19 changed files with 6204 additions and 0 deletions

View File

@ -0,0 +1,36 @@
[save_wf_after_selection]
type: logical
doc: If true, saves the wave function after the selection, before the diagonalization
interface: ezfio,provider,ocaml
default: False
[seniority_max]
type: integer
doc: Maximum number of allowed open shells. Using -1 selects all determinants
interface: ezfio,ocaml,provider
default: -1
[excitation_ref]
type: integer
doc: 1: Hartree-Fock determinant, 2:All determinants of the dominant configuration
interface: ezfio,ocaml,provider
default: 1
[excitation_max]
type: integer
doc: Maximum number of excitation with respect to the Hartree-Fock determinant. Using -1 selects all determinants
interface: ezfio,ocaml,provider
default: -1
[excitation_alpha_max]
type: integer
doc: Maximum number of excitation for alpha determinants with respect to the Hartree-Fock determinant. Using -1 selects all determinants
interface: ezfio,ocaml,provider
default: -1
[excitation_beta_max]
type: integer
doc: Maximum number of excitation for beta determinants with respect to the Hartree-Fock determinant. Using -1 selects all determinants
interface: ezfio,ocaml,provider
default: -1

View File

@ -0,0 +1,6 @@
mpi
perturbation
zmq
iterations_tc
csf
tc_bi_ortho

View File

@ -0,0 +1,136 @@
subroutine run_cipsi
BEGIN_DOC
! Selected Full Configuration Interaction with deterministic selection and
! stochastic PT2.
END_DOC
use selection_types
implicit none
integer :: i,j,k,ndet
type(pt2_type) :: pt2_data, pt2_data_err
double precision, allocatable :: zeros(:)
integer :: to_select
logical, external :: qp_stop
double precision :: threshold_generators_save
double precision :: rss
double precision, external :: memory_of_double
double precision :: correlation_energy_ratio,E_denom,E_tc,norm
PROVIDE H_apply_buffer_allocated distributed_davidson
print*,'Diagonal elements of the Fock matrix '
do i = 1, mo_num
write(*,*)i,Fock_matrix_tc_mo_tot(i,i)
enddo
N_iter = 1
threshold_generators = 1.d0
SOFT_TOUCH threshold_generators
rss = memory_of_double(N_states)*4.d0
call check_mem(rss,irp_here)
allocate (zeros(N_states))
call pt2_alloc(pt2_data, N_states)
call pt2_alloc(pt2_data_err, N_states)
double precision :: hf_energy_ref
logical :: has, print_pt2
double precision :: relative_error
relative_error=PT2_relative_error
zeros = 0.d0
pt2_data % pt2 = -huge(1.e0)
pt2_data % rpt2 = -huge(1.e0)
pt2_data % overlap(:,:) = 0.d0
pt2_data % variance = huge(1.e0)
if (s2_eig) then
call make_s2_eigenfunction
endif
print_pt2 = .False.
call diagonalize_CI_tc_bi_ortho(ndet, E_tc,norm,pt2_data,print_pt2)
call ezfio_has_hartree_fock_energy(has)
if (has) then
call ezfio_get_hartree_fock_energy(hf_energy_ref)
else
hf_energy_ref = ref_bitmask_energy
endif
if (N_det > N_det_max) then
psi_det(1:N_int,1:2,1:N_det) = psi_det_sorted_tc_gen(1:N_int,1:2,1:N_det)
psi_coef(1:N_det,1:N_states) = psi_coef_sorted_tc_gen(1:N_det,1:N_states)
N_det = N_det_max
soft_touch N_det psi_det psi_coef
if (s2_eig) then
call make_s2_eigenfunction
endif
print_pt2 = .False.
call diagonalize_CI_tc_bi_ortho(ndet, E_tc,norm,pt2_data,print_pt2)
! call routine_save_right
endif
correlation_energy_ratio = 0.d0
print_pt2 = .True.
do while ( &
(N_det < N_det_max) .and. &
(maxval(abs(pt2_data % pt2(1:N_states))) > pt2_max) &
)
write(*,'(A)') '--------------------------------------------------------------------------------'
to_select = int(sqrt(dble(N_states))*dble(N_det)*selection_factor)
to_select = max(N_states_diag, to_select)
E_denom = E_tc ! TC Energy of the current wave function
if (do_pt2) then
call pt2_dealloc(pt2_data)
call pt2_dealloc(pt2_data_err)
call pt2_alloc(pt2_data, N_states)
call pt2_alloc(pt2_data_err, N_states)
threshold_generators_save = threshold_generators
threshold_generators = 1.d0
SOFT_TOUCH threshold_generators
call ZMQ_pt2(E_denom, pt2_data, pt2_data_err, relative_error,to_select) ! Stochastic PT2 and selection
threshold_generators = threshold_generators_save
SOFT_TOUCH threshold_generators
else
call pt2_dealloc(pt2_data)
call pt2_alloc(pt2_data, N_states)
call ZMQ_selection(to_select, pt2_data)
endif
N_iter += 1
if (qp_stop()) exit
! Add selected determinants
call copy_H_apply_buffer_to_wf()
if (save_wf_after_selection) then
call save_wavefunction
endif
PROVIDE psi_coef
PROVIDE psi_det
PROVIDE psi_det_sorted_tc
call diagonalize_CI_tc_bi_ortho(ndet, E_tc,norm,pt2_data,print_pt2)
if (qp_stop()) exit
enddo
call pt2_dealloc(pt2_data)
call pt2_dealloc(pt2_data_err)
call pt2_alloc(pt2_data, N_states)
call pt2_alloc(pt2_data_err, N_states)
call ZMQ_pt2(E_tc, pt2_data, pt2_data_err, relative_error,0) ! Stochastic PT2 and selection
call diagonalize_CI_tc_bi_ortho(ndet, E_tc,norm,pt2_data,print_pt2)
end

View File

@ -0,0 +1,51 @@
BEGIN_PROVIDER [ logical, initialize_pt2_E0_denominator ]
implicit none
BEGIN_DOC
! If true, initialize pt2_E0_denominator
END_DOC
initialize_pt2_E0_denominator = .True.
END_PROVIDER
BEGIN_PROVIDER [ double precision, pt2_E0_denominator, (N_states) ]
implicit none
BEGIN_DOC
! E0 in the denominator of the PT2
END_DOC
integer :: i,j
pt2_E0_denominator = eigval_right_tc_bi_orth
! if (initialize_pt2_E0_denominator) then
! if (h0_type == "EN") then
! pt2_E0_denominator(1:N_states) = psi_energy(1:N_states)
! else if (h0_type == "HF") then
! do i=1,N_states
! j = maxloc(abs(psi_coef(:,i)),1)
! pt2_E0_denominator(i) = psi_det_hii(j)
! enddo
! else if (h0_type == "Barycentric") then
! pt2_E0_denominator(1:N_states) = barycentric_electronic_energy(1:N_states)
! else if (h0_type == "CFG") then
! pt2_E0_denominator(1:N_states) = psi_energy(1:N_states)
! else
! print *, h0_type, ' not implemented'
! stop
! endif
! do i=1,N_states
! call write_double(6,pt2_E0_denominator(i)+nuclear_repulsion, 'PT2 Energy denominator')
! enddo
! else
! pt2_E0_denominator = -huge(1.d0)
! endif
END_PROVIDER
BEGIN_PROVIDER [ double precision, pt2_overlap, (N_states, N_states) ]
implicit none
BEGIN_DOC
! Overlap between the perturbed wave functions
END_DOC
pt2_overlap(1:N_states,1:N_states) = 0.d0
END_PROVIDER

View File

@ -0,0 +1,14 @@
BEGIN_PROVIDER [ integer, nthreads_pt2 ]
implicit none
BEGIN_DOC
! Number of threads for Davidson
END_DOC
nthreads_pt2 = nproc
character*(32) :: env
call getenv('QP_NTHREADS_PT2',env)
if (trim(env) /= '') then
read(env,*) nthreads_pt2
call write_int(6,nthreads_pt2,'Target number of threads for PT2')
endif
END_PROVIDER

File diff suppressed because it is too large Load Diff

View File

View File

@ -0,0 +1,89 @@
subroutine pt2_tc_bi_ortho
use selection_types
implicit none
BEGIN_DOC
! Selected Full Configuration Interaction with Stochastic selection and PT2.
END_DOC
integer :: i,j,k,ndet
double precision, allocatable :: zeros(:)
integer :: to_select
type(pt2_type) :: pt2_data, pt2_data_err
logical, external :: qp_stop
logical :: print_pt2
double precision :: rss
double precision, external :: memory_of_double
double precision :: correlation_energy_ratio,E_denom,E_tc,norm
double precision, allocatable :: ept2(:), pt1(:),extrap_energy(:)
PROVIDE H_apply_buffer_allocated distributed_davidson mo_two_e_integrals_in_map
print*,'Diagonal elements of the Fock matrix '
do i = 1, mo_num
write(*,*)i,Fock_matrix_tc_mo_tot(i,i)
enddo
N_iter = 1
threshold_generators = 1.d0
SOFT_TOUCH threshold_generators
rss = memory_of_double(N_states)*4.d0
call check_mem(rss,irp_here)
allocate (zeros(N_states))
call pt2_alloc(pt2_data, N_states)
call pt2_alloc(pt2_data_err, N_states)
double precision :: hf_energy_ref
logical :: has
double precision :: relative_error
relative_error=PT2_relative_error
zeros = 0.d0
pt2_data % pt2 = -huge(1.e0)
pt2_data % rpt2 = -huge(1.e0)
pt2_data % overlap= 0.d0
pt2_data % variance = huge(1.e0)
if (s2_eig) then
call make_s2_eigenfunction
endif
print_pt2 = .False.
call diagonalize_CI_tc_bi_ortho(ndet, E_tc,norm,pt2_data,print_pt2)
! call routine_save_right
if (N_det > N_det_max) then
psi_det(1:N_int,1:2,1:N_det) = psi_det_sorted_tc_gen(1:N_int,1:2,1:N_det)
psi_coef(1:N_det,1:N_states) = psi_coef_sorted_tc_gen(1:N_det,1:N_states)
N_det = N_det_max
soft_touch N_det psi_det psi_coef
if (s2_eig) then
call make_s2_eigenfunction
endif
print_pt2 = .False.
call diagonalize_CI_tc_bi_ortho(ndet, E_tc,norm,pt2_data,print_pt2)
endif
allocate(ept2(1000),pt1(1000),extrap_energy(100))
correlation_energy_ratio = 0.d0
! thresh_it_dav = 5.d-5
! soft_touch thresh_it_dav
print_pt2 = .True.
to_select = int(sqrt(dble(N_states))*dble(N_det)*selection_factor)
to_select = max(N_states_diag, to_select)
E_denom = E_tc ! TC Energy of the current wave function
call pt2_dealloc(pt2_data)
call pt2_dealloc(pt2_data_err)
call pt2_alloc(pt2_data, N_states)
call pt2_alloc(pt2_data_err, N_states)
call ZMQ_pt2(E_denom, pt2_data, pt2_data_err, relative_error,to_select) ! Stochastic PT2 and selection
N_iter += 1
call diagonalize_CI_tc_bi_ortho(ndet, E_tc,norm,pt2_data,print_pt2)
end

View File

@ -0,0 +1,869 @@
BEGIN_PROVIDER [ integer, pt2_stoch_istate ]
implicit none
BEGIN_DOC
! State for stochatsic PT2
END_DOC
pt2_stoch_istate = 1
END_PROVIDER
BEGIN_PROVIDER [ integer, pt2_F, (N_det_generators) ]
&BEGIN_PROVIDER [ integer, pt2_n_tasks_max ]
implicit none
logical, external :: testTeethBuilding
integer :: i,j
pt2_n_tasks_max = elec_alpha_num*elec_alpha_num + elec_alpha_num*elec_beta_num - n_core_orb*2
pt2_n_tasks_max = min(pt2_n_tasks_max,1+N_det_generators/10000)
call write_int(6,pt2_n_tasks_max,'pt2_n_tasks_max')
pt2_F(:) = max(int(sqrt(float(pt2_n_tasks_max))),1)
do i=1,pt2_n_0(1+pt2_N_teeth/4)
pt2_F(i) = pt2_n_tasks_max*pt2_min_parallel_tasks
enddo
do i=1+pt2_n_0(pt2_N_teeth-pt2_N_teeth/4), pt2_n_0(pt2_N_teeth-pt2_N_teeth/10)
pt2_F(i) = pt2_min_parallel_tasks
enddo
do i=1+pt2_n_0(pt2_N_teeth-pt2_N_teeth/10), N_det_generators
pt2_F(i) = 1
enddo
END_PROVIDER
BEGIN_PROVIDER [ integer, pt2_N_teeth ]
&BEGIN_PROVIDER [ integer, pt2_minDetInFirstTeeth ]
implicit none
logical, external :: testTeethBuilding
if(N_det_generators < 500) then
pt2_minDetInFirstTeeth = 1
pt2_N_teeth = 1
else
pt2_minDetInFirstTeeth = min(5, N_det_generators)
do pt2_N_teeth=100,2,-1
if(testTeethBuilding(pt2_minDetInFirstTeeth, pt2_N_teeth)) exit
end do
end if
call write_int(6,pt2_N_teeth,'Number of comb teeth')
END_PROVIDER
logical function testTeethBuilding(minF, N)
implicit none
integer, intent(in) :: minF, N
integer :: n0, i
double precision :: u0, Wt, r
double precision, allocatable :: tilde_w(:), tilde_cW(:)
integer, external :: dress_find_sample
double precision :: rss
double precision, external :: memory_of_double, memory_of_int
rss = memory_of_double(2*N_det_generators+1)
call check_mem(rss,irp_here)
allocate(tilde_w(N_det_generators), tilde_cW(0:N_det_generators))
double precision :: norm2
norm2 = 0.d0
do i=N_det_generators,1,-1
tilde_w(i) = psi_coef_sorted_tc_gen(i,pt2_stoch_istate) * &
psi_coef_sorted_tc_gen(i,pt2_stoch_istate)
norm2 = norm2 + tilde_w(i)
enddo
f = 1.d0/norm2
tilde_w(:) = tilde_w(:) * f
tilde_cW(0) = -1.d0
do i=1,N_det_generators
tilde_cW(i) = tilde_cW(i-1) + tilde_w(i)
enddo
tilde_cW(:) = tilde_cW(:) + 1.d0
deallocate(tilde_w)
n0 = 0
testTeethBuilding = .false.
double precision :: f
integer :: minFN
minFN = N_det_generators - minF * N
f = 1.d0/dble(N)
do
u0 = tilde_cW(n0)
r = tilde_cW(n0 + minF)
Wt = (1d0 - u0) * f
if (dabs(Wt) <= 1.d-3) then
exit
endif
if(Wt >= r - u0) then
testTeethBuilding = .true.
exit
end if
n0 += 1
if(n0 > minFN) then
exit
end if
end do
deallocate(tilde_cW)
end function
subroutine ZMQ_pt2(E, pt2_data, pt2_data_err, relative_error, N_in)
use f77_zmq
use selection_types
implicit none
integer(ZMQ_PTR) :: zmq_to_qp_run_socket, zmq_socket_pull
integer, intent(in) :: N_in
! integer, intent(inout) :: N_in
double precision, intent(in) :: relative_error, E(N_states)
type(pt2_type), intent(inout) :: pt2_data, pt2_data_err
!
integer :: i, N
double precision :: state_average_weight_save(N_states), w(N_states,4)
integer(ZMQ_PTR), external :: new_zmq_to_qp_run_socket
type(selection_buffer) :: b
PROVIDE psi_bilinear_matrix_columns_loc psi_det_alpha_unique psi_det_beta_unique
PROVIDE psi_bilinear_matrix_rows psi_det_sorted_tc_order psi_bilinear_matrix_order
PROVIDE psi_bilinear_matrix_transp_rows_loc psi_bilinear_matrix_transp_columns
PROVIDE psi_bilinear_matrix_transp_order psi_selectors_coef_transp psi_det_sorted_tc
PROVIDE psi_det_hii selection_weight pseudo_sym
PROVIDE n_act_orb n_inact_orb n_core_orb n_virt_orb n_del_orb seniority_max
PROVIDE excitation_beta_max excitation_alpha_max excitation_max
PROVIDE psi_selectors_rcoef_bi_orth_transp psi_selectors_lcoef_bi_orth_transp
if (h0_type == 'CFG') then
PROVIDE psi_configuration_hii det_to_configuration
endif
if (N_det <= max(4,N_states) .or. pt2_N_teeth < 2) then
print*,'ZMQ_selection'
call ZMQ_selection(N_in, pt2_data)
else
print*,'else ZMQ_selection'
N = max(N_in,1) * N_states
state_average_weight_save(:) = state_average_weight(:)
if (int(N,8)*2_8 > huge(1)) then
print *, irp_here, ': integer too large'
stop -1
endif
call create_selection_buffer(N, N*2, b)
ASSERT (associated(b%det))
ASSERT (associated(b%val))
do pt2_stoch_istate=1,N_states
state_average_weight(:) = 0.d0
state_average_weight(pt2_stoch_istate) = 1.d0
TOUCH state_average_weight pt2_stoch_istate selection_weight
PROVIDE nproc pt2_F mo_two_e_integrals_in_map mo_one_e_integrals pt2_w
PROVIDE psi_selectors pt2_u pt2_J pt2_R
call new_parallel_job(zmq_to_qp_run_socket, zmq_socket_pull, 'pt2')
integer, external :: zmq_put_psi
integer, external :: zmq_put_N_det_generators
integer, external :: zmq_put_N_det_selectors
integer, external :: zmq_put_dvector
integer, external :: zmq_put_ivector
if (zmq_put_psi(zmq_to_qp_run_socket,1) == -1) then
stop 'Unable to put psi on ZMQ server'
endif
if (zmq_put_N_det_generators(zmq_to_qp_run_socket, 1) == -1) then
stop 'Unable to put N_det_generators on ZMQ server'
endif
if (zmq_put_N_det_selectors(zmq_to_qp_run_socket, 1) == -1) then
stop 'Unable to put N_det_selectors on ZMQ server'
endif
if (zmq_put_dvector(zmq_to_qp_run_socket,1,'energy',pt2_e0_denominator,size(pt2_e0_denominator)) == -1) then
stop 'Unable to put energy on ZMQ server'
endif
if (zmq_put_dvector(zmq_to_qp_run_socket,1,'state_average_weight',state_average_weight,N_states) == -1) then
stop 'Unable to put state_average_weight on ZMQ server'
endif
if (zmq_put_dvector(zmq_to_qp_run_socket,1,'selection_weight',selection_weight,N_states) == -1) then
stop 'Unable to put selection_weight on ZMQ server'
endif
if (zmq_put_ivector(zmq_to_qp_run_socket,1,'pt2_stoch_istate',pt2_stoch_istate,1) == -1) then
stop 'Unable to put pt2_stoch_istate on ZMQ server'
endif
if (zmq_put_dvector(zmq_to_qp_run_socket,1,'threshold_generators',(/threshold_generators/),1) == -1) then
stop 'Unable to put threshold_generators on ZMQ server'
endif
integer, external :: add_task_to_taskserver
character(300000) :: task
integer :: j,k,ipos,ifirst
ifirst=0
ipos=0
do i=1,N_det_generators
if (pt2_F(i) > 1) then
ipos += 1
endif
enddo
call write_int(6,sum(pt2_F),'Number of tasks')
call write_int(6,ipos,'Number of fragmented tasks')
ipos=1
do i= 1, N_det_generators
do j=1,pt2_F(pt2_J(i))
write(task(ipos:ipos+30),'(I9,1X,I9,1X,I9,''|'')') j, pt2_J(i), N_in
ipos += 30
if (ipos > 300000-30) then
if (add_task_to_taskserver(zmq_to_qp_run_socket,trim(task(1:ipos))) == -1) then
stop 'Unable to add task to task server'
endif
ipos=1
if (ifirst == 0) then
ifirst=1
if (zmq_set_running(zmq_to_qp_run_socket) == -1) then
print *, irp_here, ': Failed in zmq_set_running'
endif
endif
endif
end do
enddo
if (ipos > 1) then
if (add_task_to_taskserver(zmq_to_qp_run_socket,trim(task(1:ipos))) == -1) then
stop 'Unable to add task to task server'
endif
endif
integer, external :: zmq_set_running
if (zmq_set_running(zmq_to_qp_run_socket) == -1) then
print *, irp_here, ': Failed in zmq_set_running'
endif
double precision :: mem_collector, mem, rss
call resident_memory(rss)
mem_collector = 8.d0 * & ! bytes
( 1.d0*pt2_n_tasks_max & ! task_id, index
+ 0.635d0*N_det_generators & ! f,d
+ pt2_n_tasks_max*pt2_type_size(N_states) & ! pt2_data_task
+ N_det_generators*pt2_type_size(N_states) & ! pt2_data_I
+ 4.d0*(pt2_N_teeth+1) & ! S, S2, T2, T3
+ 1.d0*(N_int*2.d0*N + N) & ! selection buffer
+ 1.d0*(N_int*2.d0*N + N) & ! sort selection buffer
) / 1024.d0**3
integer :: nproc_target, ii
nproc_target = nthreads_pt2
ii = min(N_det, (elec_alpha_num*(mo_num-elec_alpha_num))**2)
do
mem = mem_collector + & !
nproc_target * 8.d0 * & ! bytes
( 0.5d0*pt2_n_tasks_max & ! task_id
+ 64.d0*pt2_n_tasks_max & ! task
+ pt2_type_size(N_states)*pt2_n_tasks_max*N_states & ! pt2, variance, overlap
+ 1.d0*pt2_n_tasks_max & ! i_generator, subset
+ 1.d0*(N_int*2.d0*ii+ ii) & ! selection buffer
+ 1.d0*(N_int*2.d0*ii+ ii) & ! sort selection buffer
+ 2.0d0*(ii) & ! preinteresting, interesting,
! prefullinteresting, fullinteresting
+ 2.0d0*(N_int*2*ii) & ! minilist, fullminilist
+ 1.0d0*(N_states*mo_num*mo_num) & ! mat
) / 1024.d0**3
if (nproc_target == 0) then
call check_mem(mem,irp_here)
nproc_target = 1
exit
endif
if (mem+rss < qp_max_mem) then
exit
endif
nproc_target = nproc_target - 1
enddo
call write_int(6,nproc_target,'Number of threads for PT2')
call write_double(6,mem,'Memory (Gb)')
call omp_set_max_active_levels(1)
print '(A)', '========== ======================= ===================== ===================== ==========='
print '(A)', ' Samples Energy Variance Norm^2 Seconds'
print '(A)', '========== ======================= ===================== ===================== ==========='
PROVIDE global_selection_buffer
!$OMP PARALLEL DEFAULT(shared) NUM_THREADS(nproc_target+1) &
!$OMP PRIVATE(i)
i = omp_get_thread_num()
if (i==0) then
call pt2_collector(zmq_socket_pull, E(pt2_stoch_istate),relative_error, pt2_data, pt2_data_err, b, N)
pt2_data % rpt2(pt2_stoch_istate) = &
pt2_data % pt2(pt2_stoch_istate)/(1.d0+pt2_data % overlap(pt2_stoch_istate,pt2_stoch_istate))
!TODO : We should use here the correct formula for the error of X/Y
pt2_data_err % rpt2(pt2_stoch_istate) = &
pt2_data_err % pt2(pt2_stoch_istate)/(1.d0 + pt2_data % overlap(pt2_stoch_istate,pt2_stoch_istate))
else
call pt2_slave_inproc(i)
endif
!$OMP END PARALLEL
call end_parallel_job(zmq_to_qp_run_socket, zmq_socket_pull, 'pt2')
call omp_set_max_active_levels(8)
print '(A)', '========== ======================= ===================== ===================== ==========='
do k=1,N_states
pt2_overlap(pt2_stoch_istate,k) = pt2_data % overlap(k,pt2_stoch_istate)
enddo
SOFT_TOUCH pt2_overlap
enddo
FREE pt2_stoch_istate
! Symmetrize overlap
do j=2,N_states
do i=1,j-1
pt2_overlap(i,j) = 0.5d0 * (pt2_overlap(i,j) + pt2_overlap(j,i))
pt2_overlap(j,i) = pt2_overlap(i,j)
enddo
enddo
print *, 'Overlap of perturbed states:'
do k=1,N_states
print *, pt2_overlap(k,:)
enddo
print *, '-------'
if (N_in > 0) then
b%cur = min(N_in,b%cur)
if (s2_eig) then
call make_selection_buffer_s2(b)
else
call remove_duplicates_in_selection_buffer(b)
endif
call fill_H_apply_buffer_no_selection(b%cur,b%det,N_int,0)
endif
call delete_selection_buffer(b)
state_average_weight(:) = state_average_weight_save(:)
TOUCH state_average_weight
call update_pt2_and_variance_weights(pt2_data, N_states)
endif
end subroutine
subroutine pt2_slave_inproc(i)
implicit none
integer, intent(in) :: i
PROVIDE global_selection_buffer
call run_pt2_slave(1,i,pt2_e0_denominator)
end
subroutine pt2_collector(zmq_socket_pull, E, relative_error, pt2_data, pt2_data_err, b, N_)
use f77_zmq
use selection_types
use bitmasks
implicit none
integer(ZMQ_PTR), intent(in) :: zmq_socket_pull
double precision, intent(in) :: relative_error, E
type(pt2_type), intent(inout) :: pt2_data, pt2_data_err
type(selection_buffer), intent(inout) :: b
integer, intent(in) :: N_
type(pt2_type), allocatable :: pt2_data_task(:)
type(pt2_type), allocatable :: pt2_data_I(:)
type(pt2_type), allocatable :: pt2_data_S(:)
type(pt2_type), allocatable :: pt2_data_S2(:)
type(pt2_type) :: pt2_data_teeth
integer(ZMQ_PTR),external :: new_zmq_to_qp_run_socket
integer(ZMQ_PTR) :: zmq_to_qp_run_socket
integer, external :: zmq_delete_tasks_async_send
integer, external :: zmq_delete_tasks_async_recv
integer, external :: zmq_abort
integer, external :: pt2_find_sample_lr
PROVIDE pt2_stoch_istate
integer :: more, n, i, p, c, t, n_tasks, U
integer, allocatable :: task_id(:)
integer, allocatable :: index(:)
double precision :: v, x, x2, x3, avg, avg2, avg3(N_states), eqt, E0, v0, n0(N_states)
double precision :: eqta(N_states)
double precision :: time, time1, time0
integer, allocatable :: f(:)
logical, allocatable :: d(:)
logical :: do_exit, stop_now, sending
logical, external :: qp_stop
type(selection_buffer) :: b2
double precision :: rss
double precision, external :: memory_of_double, memory_of_int
sending =.False.
rss = memory_of_int(pt2_n_tasks_max*2+N_det_generators*2)
rss += memory_of_double(N_states*N_det_generators)*3.d0
rss += memory_of_double(N_states*pt2_n_tasks_max)*3.d0
rss += memory_of_double(pt2_N_teeth+1)*4.d0
call check_mem(rss,irp_here)
! If an allocation is added here, the estimate of the memory should also be
! updated in ZMQ_pt2
allocate(task_id(pt2_n_tasks_max), index(pt2_n_tasks_max), f(N_det_generators))
allocate(d(N_det_generators+1))
allocate(pt2_data_task(pt2_n_tasks_max))
allocate(pt2_data_I(N_det_generators))
allocate(pt2_data_S(pt2_N_teeth+1))
allocate(pt2_data_S2(pt2_N_teeth+1))
zmq_to_qp_run_socket = new_zmq_to_qp_run_socket()
call create_selection_buffer(N_, N_*2, b2)
pt2_data % pt2(pt2_stoch_istate) = -huge(1.)
pt2_data_err % pt2(pt2_stoch_istate) = huge(1.)
pt2_data % variance(pt2_stoch_istate) = huge(1.)
pt2_data_err % variance(pt2_stoch_istate) = huge(1.)
pt2_data % overlap(:,pt2_stoch_istate) = 0.d0
pt2_data_err % overlap(:,pt2_stoch_istate) = huge(1.)
n = 1
t = 0
U = 0
do i=1,pt2_n_tasks_max
call pt2_alloc(pt2_data_task(i),N_states)
enddo
do i=1,pt2_N_teeth+1
call pt2_alloc(pt2_data_S(i),N_states)
call pt2_alloc(pt2_data_S2(i),N_states)
enddo
do i=1,N_det_generators
call pt2_alloc(pt2_data_I(i),N_states)
enddo
f(:) = pt2_F(:)
d(:) = .false.
n_tasks = 0
E0 = E
v0 = 0.d0
n0(:) = 0.d0
more = 1
call wall_time(time0)
time1 = time0
do_exit = .false.
stop_now = .false.
do while (n <= N_det_generators)
if(f(pt2_J(n)) == 0) then
d(pt2_J(n)) = .true.
do while(d(U+1))
U += 1
end do
! Deterministic part
do while(t <= pt2_N_teeth)
if(U >= pt2_n_0(t+1)) then
t=t+1
E0 = 0.d0
v0 = 0.d0
n0(:) = 0.d0
do i=pt2_n_0(t),1,-1
E0 += pt2_data_I(i) % pt2(pt2_stoch_istate)
v0 += pt2_data_I(i) % variance(pt2_stoch_istate)
n0(:) += pt2_data_I(i) % overlap(:,pt2_stoch_istate)
end do
else
exit
end if
end do
! Add Stochastic part
c = pt2_R(n)
if(c > 0) then
call pt2_alloc(pt2_data_teeth,N_states)
do p=pt2_N_teeth, 1, -1
v = pt2_u_0 + pt2_W_T * (pt2_u(c) + dble(p-1))
i = pt2_find_sample_lr(v, pt2_cW,pt2_n_0(p),pt2_n_0(p+1))
v = pt2_W_T / pt2_w(i)
call pt2_add ( pt2_data_teeth, v, pt2_data_I(i) )
call pt2_add ( pt2_data_S(p), 1.d0, pt2_data_teeth )
call pt2_add2( pt2_data_S2(p), 1.d0, pt2_data_teeth )
enddo
call pt2_dealloc(pt2_data_teeth)
avg = E0 + pt2_data_S(t) % pt2(pt2_stoch_istate) / dble(c)
avg2 = v0 + pt2_data_S(t) % variance(pt2_stoch_istate) / dble(c)
avg3(:) = n0(:) + pt2_data_S(t) % overlap(:,pt2_stoch_istate) / dble(c)
if ((avg /= 0.d0) .or. (n == N_det_generators) ) then
do_exit = .true.
endif
if (qp_stop()) then
stop_now = .True.
endif
pt2_data % pt2(pt2_stoch_istate) = avg
pt2_data % variance(pt2_stoch_istate) = avg2
pt2_data % overlap(:,pt2_stoch_istate) = avg3(:)
call wall_time(time)
! 1/(N-1.5) : see Brugger, The American Statistician (23) 4 p. 32 (1969)
if(c > 2) then
eqt = dabs((pt2_data_S2(t) % pt2(pt2_stoch_istate) / c) - (pt2_data_S(t) % pt2(pt2_stoch_istate)/c)**2) ! dabs for numerical stability
eqt = sqrt(eqt / (dble(c) - 1.5d0))
pt2_data_err % pt2(pt2_stoch_istate) = eqt
eqt = dabs((pt2_data_S2(t) % variance(pt2_stoch_istate) / c) - (pt2_data_S(t) % variance(pt2_stoch_istate)/c)**2) ! dabs for numerical stability
eqt = sqrt(eqt / (dble(c) - 1.5d0))
pt2_data_err % variance(pt2_stoch_istate) = eqt
eqta(:) = dabs((pt2_data_S2(t) % overlap(:,pt2_stoch_istate) / c) - (pt2_data_S(t) % overlap(:,pt2_stoch_istate)/c)**2) ! dabs for numerical stability
eqta(:) = sqrt(eqta(:) / (dble(c) - 1.5d0))
pt2_data_err % overlap(:,pt2_stoch_istate) = eqta(:)
if ((time - time1 > 1.d0) .or. (n==N_det_generators)) then
time1 = time
print '(I10, X, F12.6, X, G10.3, X, F10.6, X, G10.3, X, F10.6, X, G10.3, X, F10.4)', c, &
pt2_data % pt2(pt2_stoch_istate) +E, &
pt2_data_err % pt2(pt2_stoch_istate), &
pt2_data % variance(pt2_stoch_istate), &
pt2_data_err % variance(pt2_stoch_istate), &
pt2_data % overlap(pt2_stoch_istate,pt2_stoch_istate), &
pt2_data_err % overlap(pt2_stoch_istate,pt2_stoch_istate), &
time-time0
if (stop_now .or. ( &
(do_exit .and. (dabs(pt2_data_err % pt2(pt2_stoch_istate)) / &
(1.d-20 + dabs(pt2_data % pt2(pt2_stoch_istate)) ) <= relative_error))) ) then
if (zmq_abort(zmq_to_qp_run_socket) == -1) then
call sleep(10)
if (zmq_abort(zmq_to_qp_run_socket) == -1) then
print *, irp_here, ': Error in sending abort signal (2)'
endif
endif
endif
endif
endif
end if
n += 1
else if(more == 0) then
exit
else
call pull_pt2_results(zmq_socket_pull, index, pt2_data_task, task_id, n_tasks, b2)
if(n_tasks > pt2_n_tasks_max)then
print*,'PB !!!'
print*,'If you see this, send a bug report with the following content'
print*,irp_here
print*,'n_tasks,pt2_n_tasks_max = ',n_tasks,pt2_n_tasks_max
stop -1
endif
if (zmq_delete_tasks_async_send(zmq_to_qp_run_socket,task_id,n_tasks,sending) == -1) then
stop 'PT2: Unable to delete tasks (send)'
endif
do i=1,n_tasks
if(index(i).gt.size(pt2_data_I,1).or.index(i).lt.1)then
print*,'PB !!!'
print*,'If you see this, send a bug report with the following content'
print*,irp_here
print*,'i,index(i),size(pt2_data_I,1) = ',i,index(i),size(pt2_data_I,1)
stop -1
endif
call pt2_add(pt2_data_I(index(i)),1.d0,pt2_data_task(i))
f(index(i)) -= 1
end do
do i=1, b2%cur
! We assume the pulled buffer is sorted
if (b2%val(i) > b%mini) exit
call add_to_selection_buffer(b, b2%det(1,1,i), b2%val(i))
end do
if (zmq_delete_tasks_async_recv(zmq_to_qp_run_socket,more,sending) == -1) then
stop 'PT2: Unable to delete tasks (recv)'
endif
end if
end do
do i=1,N_det_generators
call pt2_dealloc(pt2_data_I(i))
enddo
do i=1,pt2_N_teeth+1
call pt2_dealloc(pt2_data_S(i))
call pt2_dealloc(pt2_data_S2(i))
enddo
do i=1,pt2_n_tasks_max
call pt2_dealloc(pt2_data_task(i))
enddo
!print *, 'deleting b2'
call delete_selection_buffer(b2)
!print *, 'sorting b'
call sort_selection_buffer(b)
!print *, 'done'
call end_zmq_to_qp_run_socket(zmq_to_qp_run_socket)
end subroutine
integer function pt2_find_sample(v, w)
implicit none
double precision, intent(in) :: v, w(0:N_det_generators)
integer, external :: pt2_find_sample_lr
pt2_find_sample = pt2_find_sample_lr(v, w, 0, N_det_generators)
end function
integer function pt2_find_sample_lr(v, w, l_in, r_in)
implicit none
double precision, intent(in) :: v, w(0:N_det_generators)
integer, intent(in) :: l_in,r_in
integer :: i,l,r
l=l_in
r=r_in
do while(r-l > 1)
i = shiftr(r+l,1)
if(w(i) < v) then
l = i
else
r = i
end if
end do
i = r
do r=i+1,N_det_generators
if (w(r) /= w(i)) then
exit
endif
enddo
pt2_find_sample_lr = r-1
end function
BEGIN_PROVIDER [ integer, pt2_n_tasks ]
implicit none
BEGIN_DOC
! Number of parallel tasks for the Monte Carlo
END_DOC
pt2_n_tasks = N_det_generators
END_PROVIDER
BEGIN_PROVIDER[ double precision, pt2_u, (N_det_generators)]
implicit none
integer, allocatable :: seed(:)
integer :: m,i
call random_seed(size=m)
allocate(seed(m))
do i=1,m
seed(i) = i
enddo
call random_seed(put=seed)
deallocate(seed)
call RANDOM_NUMBER(pt2_u)
END_PROVIDER
BEGIN_PROVIDER[ integer, pt2_J, (N_det_generators)]
&BEGIN_PROVIDER[ integer, pt2_R, (N_det_generators)]
implicit none
BEGIN_DOC
! pt2_J contains the list of generators after ordering them according to the
! Monte Carlo sampling.
!
! pt2_R(i) is the number of combs drawn when determinant i is computed.
END_DOC
integer :: N_c, N_j
integer :: U, t, i
double precision :: v
integer, external :: pt2_find_sample_lr
logical, allocatable :: pt2_d(:)
integer :: m,l,r,k
integer :: ncache
integer, allocatable :: ii(:,:)
double precision :: dt
ncache = min(N_det_generators,10000)
double precision :: rss
double precision, external :: memory_of_double, memory_of_int
rss = memory_of_int(ncache)*dble(pt2_N_teeth) + memory_of_int(N_det_generators)
call check_mem(rss,irp_here)
allocate(ii(pt2_N_teeth,ncache),pt2_d(N_det_generators))
pt2_R(:) = 0
pt2_d(:) = .false.
N_c = 0
N_j = pt2_n_0(1)
do i=1,N_j
pt2_d(i) = .true.
pt2_J(i) = i
end do
U = 0
do while(N_j < pt2_n_tasks)
if (N_c+ncache > N_det_generators) then
ncache = N_det_generators - N_c
endif
!$OMP PARALLEL DO DEFAULT(SHARED) PRIVATE(dt,v,t,k)
do k=1, ncache
dt = pt2_u_0
do t=1, pt2_N_teeth
v = dt + pt2_W_T *pt2_u(N_c+k)
dt = dt + pt2_W_T
ii(t,k) = pt2_find_sample_lr(v, pt2_cW,pt2_n_0(t),pt2_n_0(t+1))
end do
enddo
!$OMP END PARALLEL DO
do k=1,ncache
!ADD_COMB
N_c = N_c+1
do t=1, pt2_N_teeth
i = ii(t,k)
if(.not. pt2_d(i)) then
N_j += 1
pt2_J(N_j) = i
pt2_d(i) = .true.
end if
end do
pt2_R(N_j) = N_c
!FILL_TOOTH
do while(U < N_det_generators)
U += 1
if(.not. pt2_d(U)) then
N_j += 1
pt2_J(N_j) = U
pt2_d(U) = .true.
exit
end if
end do
if (N_j >= pt2_n_tasks) exit
end do
enddo
if(N_det_generators > 1) then
pt2_R(N_det_generators-1) = 0
pt2_R(N_det_generators) = N_c
end if
deallocate(ii,pt2_d)
END_PROVIDER
BEGIN_PROVIDER [ double precision, pt2_w, (N_det_generators) ]
&BEGIN_PROVIDER [ double precision, pt2_cW, (0:N_det_generators) ]
&BEGIN_PROVIDER [ double precision, pt2_W_T ]
&BEGIN_PROVIDER [ double precision, pt2_u_0 ]
&BEGIN_PROVIDER [ integer, pt2_n_0, (pt2_N_teeth+1) ]
implicit none
integer :: i, t
double precision, allocatable :: tilde_w(:), tilde_cW(:)
double precision :: r, tooth_width
integer, external :: pt2_find_sample
double precision :: rss
double precision, external :: memory_of_double, memory_of_int
rss = memory_of_double(2*N_det_generators+1)
call check_mem(rss,irp_here)
if (N_det_generators == 1) then
pt2_w(1) = 1.d0
pt2_cw(1) = 1.d0
pt2_u_0 = 1.d0
pt2_W_T = 0.d0
pt2_n_0(1) = 0
pt2_n_0(2) = 1
else
allocate(tilde_w(N_det_generators), tilde_cW(0:N_det_generators))
tilde_cW(0) = 0d0
do i=1,N_det_generators
tilde_w(i) = psi_coef_sorted_tc_gen(i,pt2_stoch_istate)**2 !+ 1.d-20
enddo
double precision :: norm2
norm2 = 0.d0
do i=N_det_generators,1,-1
norm2 += tilde_w(i)
enddo
tilde_w(:) = tilde_w(:) / norm2
tilde_cW(0) = -1.d0
do i=1,N_det_generators
tilde_cW(i) = tilde_cW(i-1) + tilde_w(i)
enddo
tilde_cW(:) = tilde_cW(:) + 1.d0
pt2_n_0(1) = 0
do
pt2_u_0 = tilde_cW(pt2_n_0(1))
r = tilde_cW(pt2_n_0(1) + pt2_minDetInFirstTeeth)
pt2_W_T = (1d0 - pt2_u_0) / dble(pt2_N_teeth)
if(pt2_W_T >= r - pt2_u_0) then
exit
end if
pt2_n_0(1) += 1
if(N_det_generators - pt2_n_0(1) < pt2_minDetInFirstTeeth * pt2_N_teeth) then
print *, "teeth building failed"
stop -1
end if
end do
do t=2, pt2_N_teeth
r = pt2_u_0 + pt2_W_T * dble(t-1)
pt2_n_0(t) = pt2_find_sample(r, tilde_cW)
end do
pt2_n_0(pt2_N_teeth+1) = N_det_generators
pt2_w(:pt2_n_0(1)) = tilde_w(:pt2_n_0(1))
do t=1, pt2_N_teeth
tooth_width = tilde_cW(pt2_n_0(t+1)) - tilde_cW(pt2_n_0(t))
if (tooth_width == 0.d0) then
tooth_width = sum(tilde_w(pt2_n_0(t):pt2_n_0(t+1)))
endif
ASSERT(tooth_width > 0.d0)
do i=pt2_n_0(t)+1, pt2_n_0(t+1)
pt2_w(i) = tilde_w(i) * pt2_W_T / tooth_width
end do
end do
pt2_cW(0) = 0d0
do i=1,N_det_generators
pt2_cW(i) = pt2_cW(i-1) + pt2_w(i)
end do
pt2_n_0(pt2_N_teeth+1) = N_det_generators
endif
END_PROVIDER

View File

@ -0,0 +1,128 @@
subroutine pt2_alloc(pt2_data,N)
implicit none
use selection_types
type(pt2_type), intent(inout) :: pt2_data
integer, intent(in) :: N
integer :: k
allocate(pt2_data % pt2(N) &
,pt2_data % variance(N) &
,pt2_data % rpt2(N) &
,pt2_data % overlap(N,N) &
)
pt2_data % pt2(:) = 0.d0
pt2_data % variance(:) = 0.d0
pt2_data % rpt2(:) = 0.d0
pt2_data % overlap(:,:) = 0.d0
end subroutine
subroutine pt2_dealloc(pt2_data)
implicit none
use selection_types
type(pt2_type), intent(inout) :: pt2_data
deallocate(pt2_data % pt2 &
,pt2_data % variance &
,pt2_data % rpt2 &
,pt2_data % overlap &
)
end subroutine
subroutine pt2_add(p1, w, p2)
implicit none
use selection_types
BEGIN_DOC
! p1 += w * p2
END_DOC
type(pt2_type), intent(inout) :: p1
double precision, intent(in) :: w
type(pt2_type), intent(in) :: p2
if (w == 1.d0) then
p1 % pt2(:) = p1 % pt2(:) + p2 % pt2(:)
p1 % rpt2(:) = p1 % rpt2(:) + p2 % rpt2(:)
p1 % variance(:) = p1 % variance(:) + p2 % variance(:)
p1 % overlap(:,:) = p1 % overlap(:,:) + p2 % overlap(:,:)
else
p1 % pt2(:) = p1 % pt2(:) + w * p2 % pt2(:)
p1 % rpt2(:) = p1 % rpt2(:) + w * p2 % rpt2(:)
p1 % variance(:) = p1 % variance(:) + w * p2 % variance(:)
p1 % overlap(:,:) = p1 % overlap(:,:) + w * p2 % overlap(:,:)
endif
end subroutine
subroutine pt2_add2(p1, w, p2)
implicit none
use selection_types
BEGIN_DOC
! p1 += w * p2**2
END_DOC
type(pt2_type), intent(inout) :: p1
double precision, intent(in) :: w
type(pt2_type), intent(in) :: p2
if (w == 1.d0) then
p1 % pt2(:) = p1 % pt2(:) + p2 % pt2(:) * p2 % pt2(:)
p1 % rpt2(:) = p1 % rpt2(:) + p2 % rpt2(:) * p2 % rpt2(:)
p1 % variance(:) = p1 % variance(:) + p2 % variance(:) * p2 % variance(:)
p1 % overlap(:,:) = p1 % overlap(:,:) + p2 % overlap(:,:) * p2 % overlap(:,:)
else
p1 % pt2(:) = p1 % pt2(:) + w * p2 % pt2(:) * p2 % pt2(:)
p1 % rpt2(:) = p1 % rpt2(:) + w * p2 % rpt2(:) * p2 % rpt2(:)
p1 % variance(:) = p1 % variance(:) + w * p2 % variance(:) * p2 % variance(:)
p1 % overlap(:,:) = p1 % overlap(:,:) + w * p2 % overlap(:,:) * p2 % overlap(:,:)
endif
end subroutine
subroutine pt2_serialize(pt2_data, n, x)
implicit none
use selection_types
type(pt2_type), intent(in) :: pt2_data
integer, intent(in) :: n
double precision, intent(out) :: x(*)
integer :: i,k,n2
n2 = n*n
x(1:n) = pt2_data % pt2(1:n)
k=n
x(k+1:k+n) = pt2_data % rpt2(1:n)
k=k+n
x(k+1:k+n) = pt2_data % variance(1:n)
k=k+n
x(k+1:k+n2) = reshape(pt2_data % overlap(1:n,1:n), (/ n2 /))
end
subroutine pt2_deserialize(pt2_data, n, x)
implicit none
use selection_types
type(pt2_type), intent(inout) :: pt2_data
integer, intent(in) :: n
double precision, intent(in) :: x(*)
integer :: i,k,n2
n2 = n*n
pt2_data % pt2(1:n) = x(1:n)
k=n
pt2_data % rpt2(1:n) = x(k+1:k+n)
k=k+n
pt2_data % variance(1:n) = x(k+1:k+n)
k=k+n
pt2_data % overlap(1:n,1:n) = reshape(x(k+1:k+n2), (/ n, n /))
end

View File

@ -0,0 +1,549 @@
use omp_lib
use selection_types
use f77_zmq
BEGIN_PROVIDER [ integer(omp_lock_kind), global_selection_buffer_lock ]
use omp_lib
implicit none
BEGIN_DOC
! Global buffer for the OpenMP selection
END_DOC
call omp_init_lock(global_selection_buffer_lock)
END_PROVIDER
BEGIN_PROVIDER [ type(selection_buffer), global_selection_buffer ]
use omp_lib
implicit none
BEGIN_DOC
! Global buffer for the OpenMP selection
END_DOC
call omp_set_lock(global_selection_buffer_lock)
call delete_selection_buffer(global_selection_buffer)
call create_selection_buffer(N_det_generators, 2*N_det_generators, &
global_selection_buffer)
call omp_unset_lock(global_selection_buffer_lock)
END_PROVIDER
subroutine run_pt2_slave(thread,iproc,energy)
use selection_types
use f77_zmq
implicit none
double precision, intent(in) :: energy(N_states_diag)
integer, intent(in) :: thread, iproc
if (N_det > 100000 ) then
call run_pt2_slave_large(thread,iproc,energy)
else
call run_pt2_slave_small(thread,iproc,energy)
endif
end
subroutine run_pt2_slave_small(thread,iproc,energy)
use selection_types
use f77_zmq
implicit none
double precision, intent(in) :: energy(N_states_diag)
integer, intent(in) :: thread, iproc
integer :: rc, i
integer :: worker_id, ctask, ltask
character*(512), allocatable :: task(:)
integer, allocatable :: task_id(:)
integer(ZMQ_PTR),external :: new_zmq_to_qp_run_socket
integer(ZMQ_PTR) :: zmq_to_qp_run_socket
integer(ZMQ_PTR), external :: new_zmq_push_socket
integer(ZMQ_PTR) :: zmq_socket_push
type(selection_buffer) :: b
logical :: done, buffer_ready
type(pt2_type), allocatable :: pt2_data(:)
integer :: n_tasks, k, N
integer, allocatable :: i_generator(:), subset(:)
double precision, external :: memory_of_double, memory_of_int
integer :: bsize ! Size of selection buffers
allocate(task_id(pt2_n_tasks_max), task(pt2_n_tasks_max))
allocate(pt2_data(pt2_n_tasks_max), i_generator(pt2_n_tasks_max), subset(pt2_n_tasks_max))
zmq_to_qp_run_socket = new_zmq_to_qp_run_socket()
integer, external :: connect_to_taskserver
if (connect_to_taskserver(zmq_to_qp_run_socket,worker_id,thread) == -1) then
call end_zmq_to_qp_run_socket(zmq_to_qp_run_socket)
return
endif
zmq_socket_push = new_zmq_push_socket(thread)
b%N = 0
buffer_ready = .False.
n_tasks = 1
done = .False.
do while (.not.done)
n_tasks = max(1,n_tasks)
n_tasks = min(pt2_n_tasks_max,n_tasks)
integer, external :: get_tasks_from_taskserver
if (get_tasks_from_taskserver(zmq_to_qp_run_socket,worker_id, task_id, task, n_tasks) == -1) then
exit
endif
done = task_id(n_tasks) == 0
if (done) then
n_tasks = n_tasks-1
endif
if (n_tasks == 0) exit
do k=1,n_tasks
call sscanf_ddd(task(k), subset(k), i_generator(k), N)
enddo
if (b%N == 0) then
! Only first time
bsize = min(N, (elec_alpha_num * (mo_num-elec_alpha_num))**2)
call create_selection_buffer(bsize, bsize*2, b)
buffer_ready = .True.
else
ASSERT (b%N == bsize)
endif
double precision :: time0, time1
call wall_time(time0)
do k=1,n_tasks
call pt2_alloc(pt2_data(k),N_states)
b%cur = 0
call select_connected(i_generator(k),energy,pt2_data(k),b,subset(k),pt2_F(i_generator(k)))
enddo
call wall_time(time1)
integer, external :: tasks_done_to_taskserver
if (tasks_done_to_taskserver(zmq_to_qp_run_socket,worker_id,task_id,n_tasks) == -1) then
done = .true.
endif
call sort_selection_buffer(b)
call push_pt2_results(zmq_socket_push, i_generator, pt2_data, b, task_id, n_tasks)
do k=1,n_tasks
call pt2_dealloc(pt2_data(k))
enddo
b%cur=0
! ! Try to adjust n_tasks around nproc/2 seconds per job
n_tasks = min(2*n_tasks,int( dble(n_tasks * nproc/2) / (time1 - time0 + 1.d0)))
n_tasks = min(n_tasks, pt2_n_tasks_max)
! n_tasks = 1
end do
integer, external :: disconnect_from_taskserver
do i=1,300
if (disconnect_from_taskserver(zmq_to_qp_run_socket,worker_id) /= -2) exit
call usleep(500)
print *, 'Retry disconnect...'
end do
call end_zmq_push_socket(zmq_socket_push,thread)
call end_zmq_to_qp_run_socket(zmq_to_qp_run_socket)
if (buffer_ready) then
call delete_selection_buffer(b)
endif
deallocate(pt2_data)
end subroutine
subroutine run_pt2_slave_large(thread,iproc,energy)
use selection_types
use f77_zmq
implicit none
double precision, intent(in) :: energy(N_states_diag)
integer, intent(in) :: thread, iproc
integer :: rc, i
integer :: worker_id, ctask, ltask
character*(512) :: task
integer :: task_id(1)
integer(ZMQ_PTR),external :: new_zmq_to_qp_run_socket
integer(ZMQ_PTR) :: zmq_to_qp_run_socket
integer(ZMQ_PTR), external :: new_zmq_push_socket
integer(ZMQ_PTR) :: zmq_socket_push
type(selection_buffer) :: b
logical :: done, buffer_ready
type(pt2_type) :: pt2_data
integer :: n_tasks, k, N
integer :: i_generator, subset
integer :: bsize ! Size of selection buffers
logical :: sending
double precision :: time_shift
PROVIDE global_selection_buffer global_selection_buffer_lock
call random_number(time_shift)
time_shift = time_shift*15.d0
zmq_to_qp_run_socket = new_zmq_to_qp_run_socket()
integer, external :: connect_to_taskserver
if (connect_to_taskserver(zmq_to_qp_run_socket,worker_id,thread) == -1) then
call end_zmq_to_qp_run_socket(zmq_to_qp_run_socket)
return
endif
zmq_socket_push = new_zmq_push_socket(thread)
b%N = 0
buffer_ready = .False.
n_tasks = 1
sending = .False.
done = .False.
double precision :: time0, time1
call wall_time(time0)
time0 = time0+time_shift
do while (.not.done)
integer, external :: get_tasks_from_taskserver
if (get_tasks_from_taskserver(zmq_to_qp_run_socket,worker_id, task_id, task, n_tasks) == -1) then
exit
endif
done = task_id(1) == 0
if (done) then
n_tasks = n_tasks-1
endif
if (n_tasks == 0) exit
call sscanf_ddd(task, subset, i_generator, N)
if( pt2_F(i_generator) <= 0 .or. pt2_F(i_generator) > N_det ) then
print *, irp_here
stop 'bug in selection'
endif
if (b%N == 0) then
! Only first time
bsize = min(N, (elec_alpha_num * (mo_num-elec_alpha_num))**2)
call create_selection_buffer(bsize, bsize*2, b)
buffer_ready = .True.
else
ASSERT (b%N == bsize)
endif
call pt2_alloc(pt2_data,N_states)
b%cur = 0
call select_connected(i_generator,energy,pt2_data,b,subset,pt2_F(i_generator))
integer, external :: tasks_done_to_taskserver
if (tasks_done_to_taskserver(zmq_to_qp_run_socket,worker_id,task_id,n_tasks) == -1) then
done = .true.
endif
call sort_selection_buffer(b)
call wall_time(time1)
! if (time1-time0 > 15.d0) then
call omp_set_lock(global_selection_buffer_lock)
global_selection_buffer%mini = b%mini
call merge_selection_buffers(b,global_selection_buffer)
b%cur=0
call omp_unset_lock(global_selection_buffer_lock)
call wall_time(time0)
! endif
call push_pt2_results_async_recv(zmq_socket_push,b%mini,sending)
if ( iproc == 1 .or. i_generator < 100 .or. done) then
call omp_set_lock(global_selection_buffer_lock)
call push_pt2_results_async_send(zmq_socket_push, (/i_generator/), (/pt2_data/), global_selection_buffer, (/task_id/), 1,sending)
global_selection_buffer%cur = 0
call omp_unset_lock(global_selection_buffer_lock)
else
call push_pt2_results_async_send(zmq_socket_push, (/i_generator/), (/pt2_data/), b, (/task_id/), 1,sending)
endif
call pt2_dealloc(pt2_data)
end do
call push_pt2_results_async_recv(zmq_socket_push,b%mini,sending)
integer, external :: disconnect_from_taskserver
do i=1,300
if (disconnect_from_taskserver(zmq_to_qp_run_socket,worker_id) /= -2) exit
call sleep(1)
print *, 'Retry disconnect...'
end do
call end_zmq_push_socket(zmq_socket_push,thread)
call end_zmq_to_qp_run_socket(zmq_to_qp_run_socket)
if (buffer_ready) then
call delete_selection_buffer(b)
endif
FREE global_selection_buffer
end subroutine
subroutine push_pt2_results(zmq_socket_push, index, pt2_data, b, task_id, n_tasks)
use selection_types
use f77_zmq
implicit none
integer(ZMQ_PTR), intent(in) :: zmq_socket_push
type(pt2_type), intent(in) :: pt2_data(n_tasks)
integer, intent(in) :: n_tasks, index(n_tasks), task_id(n_tasks)
type(selection_buffer), intent(inout) :: b
logical :: sending
sending = .False.
call push_pt2_results_async_send(zmq_socket_push, index, pt2_data, b, task_id, n_tasks, sending)
call push_pt2_results_async_recv(zmq_socket_push, b%mini, sending)
end subroutine
subroutine push_pt2_results_async_send(zmq_socket_push, index, pt2_data, b, task_id, n_tasks, sending)
use selection_types
use f77_zmq
implicit none
integer(ZMQ_PTR), intent(in) :: zmq_socket_push
type(pt2_type), intent(in) :: pt2_data(n_tasks)
integer, intent(in) :: n_tasks, index(n_tasks), task_id(n_tasks)
type(selection_buffer), intent(inout) :: b
logical, intent(inout) :: sending
integer :: rc, i
integer*8 :: rc8
double precision, allocatable :: pt2_serialized(:,:)
if (sending) then
print *, irp_here, ': sending is true'
stop -1
endif
sending = .True.
rc = f77_zmq_send( zmq_socket_push, n_tasks, 4, ZMQ_SNDMORE)
if (rc == -1) then
print *, irp_here, ': error sending result'
stop 1
return
else if(rc /= 4) then
stop 'push'
endif
rc = f77_zmq_send( zmq_socket_push, index, 4*n_tasks, ZMQ_SNDMORE)
if (rc == -1) then
print *, irp_here, ': error sending result'
stop 2
return
else if(rc /= 4*n_tasks) then
stop 'push'
endif
allocate(pt2_serialized (pt2_type_size(N_states),n_tasks) )
do i=1,n_tasks
call pt2_serialize(pt2_data(i),N_states,pt2_serialized(1,i))
enddo
rc = f77_zmq_send( zmq_socket_push, pt2_serialized, size(pt2_serialized)*8, ZMQ_SNDMORE)
deallocate(pt2_serialized)
if (rc == -1) then
print *, irp_here, ': error sending result'
stop 3
return
else if(rc /= size(pt2_serialized)*8) then
stop 'push'
endif
rc = f77_zmq_send( zmq_socket_push, task_id, n_tasks*4, ZMQ_SNDMORE)
if (rc == -1) then
print *, irp_here, ': error sending result'
stop 6
return
else if(rc /= 4*n_tasks) then
stop 'push'
endif
if (b%cur == 0) then
rc = f77_zmq_send( zmq_socket_push, b%cur, 4, 0)
if (rc == -1) then
print *, irp_here, ': error sending result'
stop 7
return
else if(rc /= 4) then
stop 'push'
endif
else
rc = f77_zmq_send( zmq_socket_push, b%cur, 4, ZMQ_SNDMORE)
if (rc == -1) then
print *, irp_here, ': error sending result'
stop 7
return
else if(rc /= 4) then
stop 'push'
endif
rc8 = f77_zmq_send8( zmq_socket_push, b%val, 8_8*int(b%cur,8), ZMQ_SNDMORE)
if (rc8 == -1_8) then
print *, irp_here, ': error sending result'
stop 8
return
else if(rc8 /= 8_8*int(b%cur,8)) then
stop 'push'
endif
rc8 = f77_zmq_send8( zmq_socket_push, b%det, int(bit_kind*N_int*2,8)*int(b%cur,8), 0)
if (rc8 == -1_8) then
print *, irp_here, ': error sending result'
stop 9
return
else if(rc8 /= int(N_int*2*8,8)*int(b%cur,8)) then
stop 'push'
endif
endif
end subroutine
subroutine push_pt2_results_async_recv(zmq_socket_push,mini,sending)
use selection_types
use f77_zmq
implicit none
integer(ZMQ_PTR), intent(in) :: zmq_socket_push
double precision, intent(out) :: mini
logical, intent(inout) :: sending
integer :: rc
if (.not.sending) return
! Activate is zmq_socket_push is a REQ
IRP_IF ZMQ_PUSH
IRP_ELSE
character*(2) :: ok
rc = f77_zmq_recv( zmq_socket_push, ok, 2, 0)
if (rc == -1) then
print *, irp_here, ': error sending result'
stop 10
return
else if ((rc /= 2).and.(ok(1:2) /= 'ok')) then
print *, irp_here//': error in receiving ok'
stop -1
endif
rc = f77_zmq_recv( zmq_socket_push, mini, 8, 0)
if (rc == -1) then
print *, irp_here, ': error sending result'
stop 11
return
else if (rc /= 8) then
print *, irp_here//': error in receiving mini'
stop 12
endif
IRP_ENDIF
sending = .False.
end subroutine
subroutine pull_pt2_results(zmq_socket_pull, index, pt2_data, task_id, n_tasks, b)
use selection_types
use f77_zmq
implicit none
integer(ZMQ_PTR), intent(in) :: zmq_socket_pull
type(pt2_type), intent(inout) :: pt2_data(*)
type(selection_buffer), intent(inout) :: b
integer, intent(out) :: index(*)
integer, intent(out) :: n_tasks, task_id(*)
integer :: rc, rn, i
integer*8 :: rc8
double precision, allocatable :: pt2_serialized(:,:)
rc = f77_zmq_recv( zmq_socket_pull, n_tasks, 4, 0)
if (rc == -1) then
n_tasks = 1
task_id(1) = 0
else if(rc /= 4) then
stop 'pull'
endif
rc = f77_zmq_recv( zmq_socket_pull, index, 4*n_tasks, 0)
if (rc == -1) then
n_tasks = 1
task_id(1) = 0
else if(rc /= 4*n_tasks) then
stop 'pull'
endif
allocate(pt2_serialized (pt2_type_size(N_states),n_tasks) )
rc = f77_zmq_recv( zmq_socket_pull, pt2_serialized, 8*size(pt2_serialized)*n_tasks, 0)
if (rc == -1) then
n_tasks = 1
task_id(1) = 0
else if(rc /= 8*size(pt2_serialized)) then
stop 'pull'
endif
do i=1,n_tasks
call pt2_deserialize(pt2_data(i),N_states,pt2_serialized(1,i))
enddo
deallocate(pt2_serialized)
rc = f77_zmq_recv( zmq_socket_pull, task_id, n_tasks*4, 0)
if (rc == -1) then
n_tasks = 1
task_id(1) = 0
else if(rc /= 4*n_tasks) then
stop 'pull'
endif
rc = f77_zmq_recv( zmq_socket_pull, b%cur, 4, 0)
if (rc == -1) then
n_tasks = 1
task_id(1) = 0
else if(rc /= 4) then
stop 'pull'
endif
if (b%cur > 0) then
rc8 = f77_zmq_recv8( zmq_socket_pull, b%val, 8_8*int(b%cur,8), 0)
if (rc8 == -1_8) then
n_tasks = 1
task_id(1) = 0
else if(rc8 /= 8_8*int(b%cur,8)) then
stop 'pull'
endif
rc8 = f77_zmq_recv8( zmq_socket_pull, b%det, int(bit_kind*N_int*2,8)*int(b%cur,8), 0)
if (rc8 == -1_8) then
n_tasks = 1
task_id(1) = 0
else if(rc8 /= int(N_int*2*8,8)*int(b%cur,8)) then
stop 'pull'
endif
endif
! Activate is zmq_socket_pull is a REP
IRP_IF ZMQ_PUSH
IRP_ELSE
rc = f77_zmq_send( zmq_socket_pull, 'ok', 2, ZMQ_SNDMORE)
if (rc == -1) then
n_tasks = 1
task_id(1) = 0
else if (rc /= 2) then
print *, irp_here//': error in sending ok'
stop -1
endif
rc = f77_zmq_send( zmq_socket_pull, b%mini, 8, 0)
IRP_ENDIF
end subroutine

View File

@ -0,0 +1,255 @@
subroutine run_selection_slave(thread, iproc, energy)
use f77_zmq
use selection_types
implicit none
double precision, intent(in) :: energy(N_states)
integer, intent(in) :: thread, iproc
integer :: rc, i
integer :: worker_id, task_id(1), ctask, ltask
character*(512) :: task
integer(ZMQ_PTR) :: zmq_to_qp_run_socket
integer(ZMQ_PTR) :: zmq_socket_push
integer(ZMQ_PTR), external :: new_zmq_to_qp_run_socket
integer(ZMQ_PTR), external :: new_zmq_push_socket
type(selection_buffer) :: buf, buf2
type(pt2_type) :: pt2_data
logical :: done, buffer_ready
PROVIDE psi_bilinear_matrix_columns_loc psi_det_alpha_unique psi_det_beta_unique
PROVIDE psi_bilinear_matrix_rows psi_det_sorted_tc_order psi_bilinear_matrix_order
PROVIDE psi_bilinear_matrix_transp_rows_loc psi_bilinear_matrix_transp_columns
PROVIDE psi_bilinear_matrix_transp_order N_int pt2_F pseudo_sym
PROVIDE psi_selectors_coef_transp psi_det_sorted_tc weight_selection
call pt2_alloc(pt2_data,N_states)
zmq_to_qp_run_socket = new_zmq_to_qp_run_socket()
integer, external :: connect_to_taskserver
if (connect_to_taskserver(zmq_to_qp_run_socket,worker_id,thread) == -1) then
call end_zmq_to_qp_run_socket(zmq_to_qp_run_socket)
return
endif
zmq_socket_push = new_zmq_push_socket(thread)
buf%N = 0
buffer_ready = .False.
ctask = 1
do
integer, external :: get_task_from_taskserver
if (get_task_from_taskserver(zmq_to_qp_run_socket,worker_id, task_id(ctask), task) == -1) then
exit
endif
done = task_id(ctask) == 0
if (done) then
ctask = ctask - 1
else
integer :: i_generator, N, subset, bsize
call sscanf_ddd(task, subset, i_generator, N)
if(buf%N == 0) then
! Only first time
call create_selection_buffer(N, N*2, buf)
buffer_ready = .True.
else
if (N /= buf%N) then
print *, 'N=', N
print *, 'buf%N=', buf%N
print *, 'bug in ', irp_here
stop '-1'
end if
end if
call select_connected(i_generator, energy, pt2_data, buf,subset, pt2_F(i_generator))
endif
integer, external :: task_done_to_taskserver
if(done .or. ctask == size(task_id)) then
do i=1, ctask
if (task_done_to_taskserver(zmq_to_qp_run_socket,worker_id,task_id(i)) == -1) then
call usleep(100)
if (task_done_to_taskserver(zmq_to_qp_run_socket,worker_id,task_id(i)) == -1) then
ctask = 0
done = .true.
exit
endif
endif
end do
if(ctask > 0) then
call sort_selection_buffer(buf)
! call merge_selection_buffers(buf,buf2)
call push_selection_results(zmq_socket_push, pt2_data, buf, task_id(1), ctask)
call pt2_dealloc(pt2_data)
call pt2_alloc(pt2_data,N_states)
! buf%mini = buf2%mini
buf%cur = 0
end if
ctask = 0
end if
if(done) exit
ctask = ctask + 1
end do
if(ctask > 0) then
call sort_selection_buffer(buf)
! call merge_selection_buffers(buf,buf2)
call push_selection_results(zmq_socket_push, pt2_data, buf, task_id(1), ctask)
! buf%mini = buf2%mini
buf%cur = 0
end if
ctask = 0
call pt2_dealloc(pt2_data)
integer, external :: disconnect_from_taskserver
if (disconnect_from_taskserver(zmq_to_qp_run_socket,worker_id) == -1) then
continue
endif
call end_zmq_to_qp_run_socket(zmq_to_qp_run_socket)
call end_zmq_push_socket(zmq_socket_push,thread)
if (buffer_ready) then
call delete_selection_buffer(buf)
! call delete_selection_buffer(buf2)
endif
end subroutine
subroutine push_selection_results(zmq_socket_push, pt2_data, b, task_id, ntasks)
use f77_zmq
use selection_types
implicit none
integer(ZMQ_PTR), intent(in) :: zmq_socket_push
type(pt2_type), intent(in) :: pt2_data
type(selection_buffer), intent(inout) :: b
integer, intent(in) :: ntasks, task_id(*)
integer :: rc
double precision, allocatable :: pt2_serialized(:)
rc = f77_zmq_send( zmq_socket_push, b%cur, 4, ZMQ_SNDMORE)
if(rc /= 4) then
print *, 'f77_zmq_send( zmq_socket_push, b%cur, 4, ZMQ_SNDMORE)'
endif
allocate(pt2_serialized (pt2_type_size(N_states)) )
call pt2_serialize(pt2_data,N_states,pt2_serialized)
rc = f77_zmq_send( zmq_socket_push, pt2_serialized, size(pt2_serialized)*8, ZMQ_SNDMORE)
if (rc == -1) then
print *, irp_here, ': error sending result'
stop 3
return
else if(rc /= size(pt2_serialized)*8) then
stop 'push'
endif
deallocate(pt2_serialized)
if (b%cur > 0) then
rc = f77_zmq_send( zmq_socket_push, b%val(1), 8*b%cur, ZMQ_SNDMORE)
if(rc /= 8*b%cur) then
print *, 'f77_zmq_send( zmq_socket_push, b%val(1), 8*b%cur, ZMQ_SNDMORE)'
endif
rc = f77_zmq_send( zmq_socket_push, b%det(1,1,1), bit_kind*N_int*2*b%cur, ZMQ_SNDMORE)
if(rc /= bit_kind*N_int*2*b%cur) then
print *, 'f77_zmq_send( zmq_socket_push, b%det(1,1,1), bit_kind*N_int*2*b%cur, ZMQ_SNDMORE)'
endif
endif
rc = f77_zmq_send( zmq_socket_push, ntasks, 4, ZMQ_SNDMORE)
if(rc /= 4) then
print *, 'f77_zmq_send( zmq_socket_push, ntasks, 4, ZMQ_SNDMORE)'
endif
rc = f77_zmq_send( zmq_socket_push, task_id(1), ntasks*4, 0)
if(rc /= 4*ntasks) then
print *, 'f77_zmq_send( zmq_socket_push, task_id(1), ntasks*4, 0)'
endif
! Activate is zmq_socket_push is a REQ
IRP_IF ZMQ_PUSH
IRP_ELSE
character*(2) :: ok
rc = f77_zmq_recv( zmq_socket_push, ok, 2, 0)
if ((rc /= 2).and.(ok(1:2) /= 'ok')) then
print *, irp_here//': error in receiving ok'
stop -1
endif
IRP_ENDIF
end subroutine
subroutine pull_selection_results(zmq_socket_pull, pt2_data, val, det, N, task_id, ntasks)
use f77_zmq
use selection_types
implicit none
integer(ZMQ_PTR), intent(in) :: zmq_socket_pull
type(pt2_type), intent(inout) :: pt2_data
double precision, intent(out) :: val(*)
integer(bit_kind), intent(out) :: det(N_int, 2, *)
integer, intent(out) :: N, ntasks, task_id(*)
integer :: rc, rn, i
double precision, allocatable :: pt2_serialized(:)
rc = f77_zmq_recv( zmq_socket_pull, N, 4, 0)
if(rc /= 4) then
print *, 'f77_zmq_recv( zmq_socket_pull, N, 4, 0)'
endif
allocate(pt2_serialized (pt2_type_size(N_states)) )
rc = f77_zmq_recv( zmq_socket_pull, pt2_serialized, 8*size(pt2_serialized), 0)
if (rc == -1) then
ntasks = 1
task_id(1) = 0
else if(rc /= 8*size(pt2_serialized)) then
stop 'pull'
endif
call pt2_deserialize(pt2_data,N_states,pt2_serialized)
deallocate(pt2_serialized)
if (N>0) then
rc = f77_zmq_recv( zmq_socket_pull, val(1), 8*N, 0)
if(rc /= 8*N) then
print *, 'f77_zmq_recv( zmq_socket_pull, val(1), 8*N, 0)'
endif
rc = f77_zmq_recv( zmq_socket_pull, det(1,1,1), bit_kind*N_int*2*N, 0)
if(rc /= bit_kind*N_int*2*N) then
print *, 'f77_zmq_recv( zmq_socket_pull, det(1,1,1), bit_kind*N_int*2*N, 0)'
endif
endif
rc = f77_zmq_recv( zmq_socket_pull, ntasks, 4, 0)
if(rc /= 4) then
print *, 'f77_zmq_recv( zmq_socket_pull, ntasks, 4, 0)'
endif
rc = f77_zmq_recv( zmq_socket_pull, task_id(1), ntasks*4, 0)
if(rc /= 4*ntasks) then
print *, 'f77_zmq_recv( zmq_socket_pull, task_id(1), ntasks*4, 0)'
endif
! Activate is zmq_socket_pull is a REP
IRP_IF ZMQ_PUSH
IRP_ELSE
rc = f77_zmq_send( zmq_socket_pull, 'ok', 2, 0)
if (rc /= 2) then
print *, irp_here//': error in sending ok'
stop -1
endif
IRP_ENDIF
end subroutine

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,416 @@
subroutine create_selection_buffer(N, size_in, res)
use selection_types
implicit none
BEGIN_DOC
! Allocates the memory for a selection buffer.
! The arrays have dimension size_in and the maximum number of elements is N
END_DOC
integer, intent(in) :: N, size_in
type(selection_buffer), intent(out) :: res
integer :: siz
siz = max(size_in,1)
double precision :: rss
double precision, external :: memory_of_double
rss = memory_of_double(siz)*(N_int*2+1)
call check_mem(rss,irp_here)
allocate(res%det(N_int, 2, siz), res%val(siz))
res%val(:) = 0d0
res%det(:,:,:) = 0_8
res%N = N
res%mini = 0d0
res%cur = 0
end subroutine
subroutine delete_selection_buffer(b)
use selection_types
implicit none
type(selection_buffer), intent(inout) :: b
if (associated(b%det)) then
deallocate(b%det)
endif
if (associated(b%val)) then
deallocate(b%val)
endif
NULLIFY(b%det)
NULLIFY(b%val)
b%cur = 0
b%mini = 0.d0
b%N = 0
end
subroutine add_to_selection_buffer(b, det, val)
use selection_types
implicit none
type(selection_buffer), intent(inout) :: b
integer(bit_kind), intent(in) :: det(N_int, 2)
double precision, intent(in) :: val
integer :: i
if(b%N > 0 .and. val <= b%mini) then
b%cur += 1
b%det(1:N_int,1:2,b%cur) = det(1:N_int,1:2)
b%val(b%cur) = val
if(b%cur == size(b%val)) then
call sort_selection_buffer(b)
end if
end if
end subroutine
subroutine merge_selection_buffers(b1, b2)
use selection_types
implicit none
BEGIN_DOC
! Merges the selection buffers b1 and b2 into b2
END_DOC
type(selection_buffer), intent(inout) :: b1
type(selection_buffer), intent(inout) :: b2
integer(bit_kind), pointer :: detmp(:,:,:)
double precision, pointer :: val(:)
integer :: i, i1, i2, k, nmwen, sze
if (b1%cur == 0) return
do while (b1%val(b1%cur) > b2%mini)
b1%cur = b1%cur-1
if (b1%cur == 0) then
return
endif
enddo
nmwen = min(b1%N, b1%cur+b2%cur)
double precision :: rss
double precision, external :: memory_of_double
sze = max(size(b1%val), size(b2%val))
rss = memory_of_double(sze) + 2*N_int*memory_of_double(sze)
call check_mem(rss,irp_here)
allocate(val(sze), detmp(N_int, 2, sze))
i1=1
i2=1
do i=1,nmwen
if ( (i1 > b1%cur).and.(i2 > b2%cur) ) then
exit
else if (i1 > b1%cur) then
val(i) = b2%val(i2)
detmp(1:N_int,1,i) = b2%det(1:N_int,1,i2)
detmp(1:N_int,2,i) = b2%det(1:N_int,2,i2)
i2=i2+1
else if (i2 > b2%cur) then
val(i) = b1%val(i1)
detmp(1:N_int,1,i) = b1%det(1:N_int,1,i1)
detmp(1:N_int,2,i) = b1%det(1:N_int,2,i1)
i1=i1+1
else
if (b1%val(i1) <= b2%val(i2)) then
val(i) = b1%val(i1)
detmp(1:N_int,1,i) = b1%det(1:N_int,1,i1)
detmp(1:N_int,2,i) = b1%det(1:N_int,2,i1)
i1=i1+1
else
val(i) = b2%val(i2)
detmp(1:N_int,1,i) = b2%det(1:N_int,1,i2)
detmp(1:N_int,2,i) = b2%det(1:N_int,2,i2)
i2=i2+1
endif
endif
enddo
deallocate(b2%det, b2%val)
do i=nmwen+1,b2%N
val(i) = 0.d0
detmp(1:N_int,1:2,i) = 0_bit_kind
enddo
b2%det => detmp
b2%val => val
b2%mini = min(b2%mini,b2%val(b2%N))
b2%cur = nmwen
end
subroutine sort_selection_buffer(b)
use selection_types
implicit none
type(selection_buffer), intent(inout) :: b
integer, allocatable :: iorder(:)
integer(bit_kind), pointer :: detmp(:,:,:)
integer :: i, nmwen
logical, external :: detEq
if (b%N == 0 .or. b%cur == 0) return
nmwen = min(b%N, b%cur)
double precision :: rss
double precision, external :: memory_of_double, memory_of_int
rss = memory_of_int(b%cur) + 2*N_int*memory_of_double(size(b%det,3))
call check_mem(rss,irp_here)
allocate(iorder(b%cur), detmp(N_int, 2, size(b%det,3)))
do i=1,b%cur
iorder(i) = i
end do
call dsort(b%val, iorder, b%cur)
do i=1, nmwen
detmp(1:N_int,1,i) = b%det(1:N_int,1,iorder(i))
detmp(1:N_int,2,i) = b%det(1:N_int,2,iorder(i))
end do
deallocate(b%det,iorder)
b%det => detmp
b%mini = min(b%mini,b%val(b%N))
b%cur = nmwen
end subroutine
subroutine make_selection_buffer_s2(b)
use selection_types
type(selection_buffer), intent(inout) :: b
integer(bit_kind), allocatable :: o(:,:,:)
double precision, allocatable :: val(:)
integer :: n_d
integer :: i,k,sze,n_alpha,j,n
logical :: dup
! Sort
integer, allocatable :: iorder(:)
integer*8, allocatable :: bit_tmp(:)
integer*8, external :: configuration_search_key
integer(bit_kind), allocatable :: tmp_array(:,:,:)
logical, allocatable :: duplicate(:)
n_d = b%cur
double precision :: rss
double precision, external :: memory_of_double
rss = (4*N_int+4)*memory_of_double(n_d)
call check_mem(rss,irp_here)
allocate(o(N_int,2,n_d), iorder(n_d), duplicate(n_d), bit_tmp(n_d), &
tmp_array(N_int,2,n_d), val(n_d) )
do i=1,n_d
do k=1,N_int
o(k,1,i) = ieor(b%det(k,1,i), b%det(k,2,i))
o(k,2,i) = iand(b%det(k,1,i), b%det(k,2,i))
enddo
iorder(i) = i
bit_tmp(i) = configuration_search_key(o(1,1,i),N_int)
enddo
deallocate(b%det)
call i8sort(bit_tmp,iorder,n_d)
do i=1,n_d
do k=1,N_int
tmp_array(k,1,i) = o(k,1,iorder(i))
tmp_array(k,2,i) = o(k,2,iorder(i))
enddo
val(i) = b%val(iorder(i))
duplicate(i) = .False.
enddo
! Find duplicates
do i=1,n_d-1
if (duplicate(i)) then
cycle
endif
j = i+1
do while (bit_tmp(j)==bit_tmp(i))
if (duplicate(j)) then
j+=1
if (j>n_d) then
exit
endif
cycle
endif
dup = .True.
do k=1,N_int
if ( (tmp_array(k,1,i) /= tmp_array(k,1,j)) &
.or. (tmp_array(k,2,i) /= tmp_array(k,2,j)) ) then
dup = .False.
exit
endif
enddo
if (dup) then
val(i) = max(val(i), val(j))
duplicate(j) = .True.
endif
j+=1
if (j>n_d) then
exit
endif
enddo
enddo
deallocate (b%val)
! Copy filtered result
integer :: n_p
n_p=0
do i=1,n_d
if (duplicate(i)) then
cycle
endif
n_p = n_p + 1
do k=1,N_int
o(k,1,n_p) = tmp_array(k,1,i)
o(k,2,n_p) = tmp_array(k,2,i)
enddo
val(n_p) = val(i)
enddo
! Sort by importance
do i=1,n_p
iorder(i) = i
end do
call dsort(val,iorder,n_p)
do i=1,n_p
do k=1,N_int
tmp_array(k,1,i) = o(k,1,iorder(i))
tmp_array(k,2,i) = o(k,2,iorder(i))
enddo
enddo
do i=1,n_p
do k=1,N_int
o(k,1,i) = tmp_array(k,1,i)
o(k,2,i) = tmp_array(k,2,i)
enddo
enddo
! Create determinants
n_d = 0
do i=1,n_p
call configuration_to_dets_size(o(1,1,i),sze,elec_alpha_num,N_int)
n_d = n_d + sze
if (n_d > b%cur) then
! if (n_d - b%cur > b%cur - n_d + sze) then
! n_d = n_d - sze
! endif
exit
endif
enddo
rss = (4*N_int+2)*memory_of_double(n_d)
call check_mem(rss,irp_here)
allocate(b%det(N_int,2,2*n_d), b%val(2*n_d))
k=1
do i=1,n_p
n=n_d
call configuration_to_dets_size(o(1,1,i),n,elec_alpha_num,N_int)
call configuration_to_dets(o(1,1,i),b%det(1,1,k),n,elec_alpha_num,N_int)
do j=k,k+n-1
b%val(j) = val(i)
enddo
k = k+n
if (k > n_d) exit
enddo
deallocate(o)
b%cur = n_d
b%N = n_d
end
subroutine remove_duplicates_in_selection_buffer(b)
use selection_types
type(selection_buffer), intent(inout) :: b
integer(bit_kind), allocatable :: o(:,:,:)
double precision, allocatable :: val(:)
integer :: n_d
integer :: i,k,sze,n_alpha,j,n
logical :: dup
! Sort
integer, allocatable :: iorder(:)
integer*8, allocatable :: bit_tmp(:)
integer*8, external :: det_search_key
integer(bit_kind), allocatable :: tmp_array(:,:,:)
logical, allocatable :: duplicate(:)
n_d = b%cur
logical :: found_duplicates
double precision :: rss
double precision, external :: memory_of_double
rss = (4*N_int+4)*memory_of_double(n_d)
call check_mem(rss,irp_here)
found_duplicates = .False.
allocate(iorder(n_d), duplicate(n_d), bit_tmp(n_d), &
tmp_array(N_int,2,n_d), val(n_d) )
do i=1,n_d
iorder(i) = i
bit_tmp(i) = det_search_key(b%det(1,1,i),N_int)
enddo
call i8sort(bit_tmp,iorder,n_d)
do i=1,n_d
do k=1,N_int
tmp_array(k,1,i) = b%det(k,1,iorder(i))
tmp_array(k,2,i) = b%det(k,2,iorder(i))
enddo
val(i) = b%val(iorder(i))
duplicate(i) = .False.
enddo
! Find duplicates
do i=1,n_d-1
if (duplicate(i)) then
cycle
endif
j = i+1
do while (bit_tmp(j)==bit_tmp(i))
if (duplicate(j)) then
j+=1
if (j>n_d) then
exit
endif
cycle
endif
dup = .True.
do k=1,N_int
if ( (tmp_array(k,1,i) /= tmp_array(k,1,j)) &
.or. (tmp_array(k,2,i) /= tmp_array(k,2,j)) ) then
dup = .False.
exit
endif
enddo
if (dup) then
duplicate(j) = .True.
found_duplicates = .True.
endif
j+=1
if (j>n_d) then
exit
endif
enddo
enddo
if (found_duplicates) then
! Copy filtered result
integer :: n_p
n_p=0
do i=1,n_d
if (duplicate(i)) then
cycle
endif
n_p = n_p + 1
do k=1,N_int
b%det(k,1,n_p) = tmp_array(k,1,i)
b%det(k,2,n_p) = tmp_array(k,2,i)
enddo
val(n_p) = val(i)
enddo
b%cur=n_p
b%N=n_p
endif
end

View File

@ -0,0 +1,25 @@
module selection_types
type selection_buffer
integer :: N, cur
integer(8) , pointer :: det(:,:,:)
double precision, pointer :: val(:)
double precision :: mini
endtype
type pt2_type
double precision, allocatable :: pt2(:)
double precision, allocatable :: rpt2(:)
double precision, allocatable :: variance(:)
double precision, allocatable :: overlap(:,:)
endtype
contains
integer function pt2_type_size(N)
implicit none
integer, intent(in) :: N
pt2_type_size = (3*n + n*n)
end function
end module

View File

@ -0,0 +1,134 @@
BEGIN_PROVIDER [ double precision, pt2_match_weight, (N_states) ]
implicit none
BEGIN_DOC
! Weights adjusted along the selection to make the PT2 contributions
! of each state coincide.
END_DOC
pt2_match_weight(:) = 1.d0
END_PROVIDER
BEGIN_PROVIDER [ double precision, variance_match_weight, (N_states) ]
implicit none
BEGIN_DOC
! Weights adjusted along the selection to make the variances
! of each state coincide.
END_DOC
variance_match_weight(:) = 1.d0
END_PROVIDER
subroutine update_pt2_and_variance_weights(pt2_data, N_st)
implicit none
use selection_types
BEGIN_DOC
! Updates the PT2- and Variance- matching weights.
END_DOC
integer, intent(in) :: N_st
type(pt2_type), intent(in) :: pt2_data
double precision :: pt2(N_st)
double precision :: variance(N_st)
double precision :: avg, element, dt, x
integer :: k
pt2(:) = pt2_data % pt2(:)
variance(:) = pt2_data % variance(:)
avg = sum(pt2(1:N_st)) / dble(N_st) + 1.d-32 ! Avoid future division by zero
dt = 8.d0 !* selection_factor
do k=1,N_st
element = exp(dt*(pt2(k)/avg - 1.d0))
element = min(2.0d0 , element)
element = max(0.5d0 , element)
pt2_match_weight(k) *= element
enddo
avg = sum(variance(1:N_st)) / dble(N_st) + 1.d-32 ! Avoid future division by zero
do k=1,N_st
element = exp(dt*(variance(k)/avg -1.d0))
element = min(2.0d0 , element)
element = max(0.5d0 , element)
variance_match_weight(k) *= element
enddo
if (N_det < 100) then
! For tiny wave functions, weights are 1.d0
pt2_match_weight(:) = 1.d0
variance_match_weight(:) = 1.d0
endif
threshold_davidson_pt2 = min(1.d-6, &
max(threshold_davidson, 1.e-1 * PT2_relative_error * minval(abs(pt2(1:N_states)))) )
SOFT_TOUCH pt2_match_weight variance_match_weight threshold_davidson_pt2
end
BEGIN_PROVIDER [ double precision, selection_weight, (N_states) ]
implicit none
BEGIN_DOC
! Weights used in the selection criterion
END_DOC
select case (weight_selection)
case (0)
print *, 'Using input weights in selection'
selection_weight(1:N_states) = c0_weight(1:N_states) * state_average_weight(1:N_states)
case (1)
print *, 'Using 1/c_max^2 weight in selection'
selection_weight(1:N_states) = c0_weight(1:N_states)
case (2)
print *, 'Using pt2-matching weight in selection'
selection_weight(1:N_states) = c0_weight(1:N_states) * pt2_match_weight(1:N_states)
print *, '# PT2 weight ', real(pt2_match_weight(:),4)
case (3)
print *, 'Using variance-matching weight in selection'
selection_weight(1:N_states) = c0_weight(1:N_states) * variance_match_weight(1:N_states)
print *, '# var weight ', real(variance_match_weight(:),4)
case (4)
print *, 'Using variance- and pt2-matching weights in selection'
selection_weight(1:N_states) = c0_weight(1:N_states) * sqrt(variance_match_weight(1:N_states) * pt2_match_weight(1:N_states))
print *, '# PT2 weight ', real(pt2_match_weight(:),4)
print *, '# var weight ', real(variance_match_weight(:),4)
case (5)
print *, 'Using variance-matching weight in selection'
selection_weight(1:N_states) = c0_weight(1:N_states) * variance_match_weight(1:N_states)
print *, '# var weight ', real(variance_match_weight(:),4)
case (6)
print *, 'Using CI coefficient-based selection'
selection_weight(1:N_states) = c0_weight(1:N_states)
case (7)
print *, 'Input weights multiplied by variance- and pt2-matching'
selection_weight(1:N_states) = c0_weight(1:N_states) * sqrt(variance_match_weight(1:N_states) * pt2_match_weight(1:N_states)) * state_average_weight(1:N_states)
print *, '# PT2 weight ', real(pt2_match_weight(:),4)
print *, '# var weight ', real(variance_match_weight(:),4)
case (8)
print *, 'Input weights multiplied by pt2-matching'
selection_weight(1:N_states) = c0_weight(1:N_states) * pt2_match_weight(1:N_states) * state_average_weight(1:N_states)
print *, '# PT2 weight ', real(pt2_match_weight(:),4)
case (9)
print *, 'Input weights multiplied by variance-matching'
selection_weight(1:N_states) = c0_weight(1:N_states) * variance_match_weight(1:N_states) * state_average_weight(1:N_states)
print *, '# var weight ', real(variance_match_weight(:),4)
end select
print *, '# Total weight ', real(selection_weight(:),4)
END_PROVIDER

View File

@ -0,0 +1,350 @@
subroutine run_slave_cipsi
BEGIN_DOC
! Helper program for distributed parallelism
END_DOC
implicit none
call omp_set_max_active_levels(1)
distributed_davidson = .False.
read_wf = .False.
SOFT_TOUCH read_wf distributed_davidson
call provide_everything
call switch_qp_run_to_master
call run_slave_main
end
subroutine provide_everything
PROVIDE H_apply_buffer_allocated mo_two_e_integrals_in_map psi_det_generators psi_coef_generators psi_det_sorted_bit psi_selectors n_det_generators n_states generators_bitmask zmq_context N_states_diag
PROVIDE psi_selectors_rcoef_bi_orth_transp psi_selectors_lcoef_bi_orth_transp
PROVIDE pt2_e0_denominator mo_num N_int ci_energy mpi_master zmq_state zmq_context
PROVIDE psi_det psi_coef threshold_generators state_average_weight
PROVIDE N_det_selectors pt2_stoch_istate N_det selection_weight pseudo_sym
end
subroutine run_slave_main
use f77_zmq
implicit none
IRP_IF MPI
include 'mpif.h'
IRP_ENDIF
integer(ZMQ_PTR), external :: new_zmq_to_qp_run_socket
integer(ZMQ_PTR) :: zmq_to_qp_run_socket
double precision :: energy(N_states)
character*(64) :: states(10)
character*(64) :: old_state
integer :: rc, i, ierr
double precision :: t0, t1
integer, external :: zmq_get_dvector, zmq_get_N_det_generators
integer, external :: zmq_get8_dvector
integer, external :: zmq_get_ivector
integer, external :: zmq_get_psi, zmq_get_N_det_selectors, zmq_get_psi_bilinear
integer, external :: zmq_get_psi_notouch
integer, external :: zmq_get_N_states_diag
zmq_context = f77_zmq_ctx_new ()
states(1) = 'selection'
states(2) = 'davidson'
states(3) = 'pt2'
old_state = 'Waiting'
zmq_to_qp_run_socket = new_zmq_to_qp_run_socket()
PROVIDE psi_det psi_coef threshold_generators state_average_weight mpi_master
PROVIDE zmq_state N_det_selectors pt2_stoch_istate N_det pt2_e0_denominator
PROVIDE N_det_generators N_states N_states_diag pt2_e0_denominator mpi_rank
IRP_IF MPI
call MPI_BARRIER(MPI_COMM_WORLD, ierr)
IRP_ENDIF
do
if (mpi_master) then
call wait_for_states(states,zmq_state,size(states))
if (zmq_state(1:64) == old_state(1:64)) then
call usleep(200)
cycle
else
old_state(1:64) = zmq_state(1:64)
endif
print *, trim(zmq_state)
endif
IRP_IF MPI_DEBUG
print *, irp_here, mpi_rank
call MPI_BARRIER(MPI_COMM_WORLD, ierr)
IRP_ENDIF
IRP_IF MPI
call MPI_BCAST (zmq_state, 128, MPI_CHARACTER, 0, MPI_COMM_WORLD, ierr)
if (ierr /= MPI_SUCCESS) then
print *, irp_here, 'error in broadcast of zmq_state'
endif
IRP_ENDIF
if(zmq_state(1:7) == 'Stopped') then
exit
endif
if (zmq_state(1:9) == 'selection') then
! Selection
! ---------
call wall_time(t0)
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_psi')
IRP_ENDIF
if (zmq_get_psi(zmq_to_qp_run_socket,1) == -1) cycle
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_dvector threshold_generators')
IRP_ENDIF
if (zmq_get_dvector(zmq_to_qp_run_socket,1,'threshold_generators',(/threshold_generators/),1) == -1) cycle
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_dvector energy')
IRP_ENDIF
if (zmq_get_dvector(zmq_to_qp_run_socket,1,'energy',energy,N_states) == -1) cycle
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_N_det_generators')
IRP_ENDIF
if (zmq_get_N_det_generators (zmq_to_qp_run_socket, 1) == -1) cycle
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_N_det_selectors')
IRP_ENDIF
if (zmq_get_N_det_selectors(zmq_to_qp_run_socket, 1) == -1) cycle
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_dvector state_average_weight')
IRP_ENDIF
if (zmq_get_dvector(zmq_to_qp_run_socket,1,'state_average_weight',state_average_weight,N_states) == -1) cycle
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_dvector selection_weight')
IRP_ENDIF
if (zmq_get_dvector(zmq_to_qp_run_socket,1,'selection_weight',selection_weight,N_states) == -1) cycle
pt2_e0_denominator(1:N_states) = energy(1:N_states)
TOUCH pt2_e0_denominator state_average_weight threshold_generators selection_weight psi_det psi_coef
if (mpi_master) then
print *, 'N_det', N_det
print *, 'N_det_generators', N_det_generators
print *, 'N_det_selectors', N_det_selectors
print *, 'pt2_e0_denominator', pt2_e0_denominator
print *, 'pt2_stoch_istate', pt2_stoch_istate
print *, 'state_average_weight', state_average_weight
print *, 'selection_weight', selection_weight
endif
call wall_time(t1)
call write_double(6,(t1-t0),'Broadcast time')
IRP_IF MPI_DEBUG
call mpi_print('Entering OpenMP section')
IRP_ENDIF
!$OMP PARALLEL PRIVATE(i)
i = omp_get_thread_num()
call run_selection_slave(0,i,energy)
!$OMP END PARALLEL
print *, mpi_rank, ': Selection done'
IRP_IF MPI
call MPI_BARRIER(MPI_COMM_WORLD, ierr)
if (ierr /= MPI_SUCCESS) then
print *, irp_here, 'error in barrier'
endif
IRP_ENDIF
call mpi_print('----------')
else if (zmq_state(1:8) == 'davidson') then
! Davidson
! --------
call wall_time(t0)
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_N_states_diag')
IRP_ENDIF
if (zmq_get_N_states_diag(zmq_to_qp_run_socket,1) == -1) cycle
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_psi')
IRP_ENDIF
if (zmq_get_psi(zmq_to_qp_run_socket,1) == -1) cycle
call wall_time(t1)
call write_double(6,(t1-t0),'Broadcast time')
!---
call omp_set_max_active_levels(8)
call davidson_slave_tcp(0)
call omp_set_max_active_levels(1)
print *, mpi_rank, ': Davidson done'
!---
IRP_IF MPI
call MPI_BARRIER(MPI_COMM_WORLD, ierr)
if (ierr /= MPI_SUCCESS) then
print *, irp_here, 'error in barrier'
endif
IRP_ENDIF
call mpi_print('----------')
else if (zmq_state(1:3) == 'pt2') then
! PT2
! ---
IRP_IF MPI
call MPI_BARRIER(MPI_COMM_WORLD, ierr)
if (ierr /= MPI_SUCCESS) then
print *, irp_here, 'error in barrier'
endif
IRP_ENDIF
call wall_time(t0)
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_psi')
IRP_ENDIF
if (zmq_get_psi(zmq_to_qp_run_socket,1) == -1) cycle
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_N_det_generators')
IRP_ENDIF
if (zmq_get_N_det_generators (zmq_to_qp_run_socket, 1) == -1) cycle
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_N_det_selectors')
IRP_ENDIF
if (zmq_get_N_det_selectors(zmq_to_qp_run_socket, 1) == -1) cycle
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_dvector threshold_generators')
IRP_ENDIF
if (zmq_get_dvector(zmq_to_qp_run_socket,1,'threshold_generators',(/threshold_generators/),1) == -1) cycle
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_dvector energy')
IRP_ENDIF
if (zmq_get_dvector(zmq_to_qp_run_socket,1,'energy',energy,N_states) == -1) cycle
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_ivector pt2_stoch_istate')
IRP_ENDIF
if (zmq_get_ivector(zmq_to_qp_run_socket,1,'pt2_stoch_istate',pt2_stoch_istate,1) == -1) cycle
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_dvector state_average_weight')
IRP_ENDIF
if (zmq_get_dvector(zmq_to_qp_run_socket,1,'state_average_weight',state_average_weight,N_states) == -1) cycle
IRP_IF MPI_DEBUG
call mpi_print('zmq_get_dvector selection_weight')
IRP_ENDIF
if (zmq_get_dvector(zmq_to_qp_run_socket,1,'selection_weight',selection_weight,N_states) == -1) cycle
pt2_e0_denominator(1:N_states) = energy(1:N_states)
SOFT_TOUCH pt2_e0_denominator state_average_weight pt2_stoch_istate threshold_generators selection_weight psi_det psi_coef N_det_generators N_det_selectors
call wall_time(t1)
call write_double(6,(t1-t0),'Broadcast time')
IRP_IF MPI
call MPI_BARRIER(MPI_COMM_WORLD, ierr)
if (ierr /= MPI_SUCCESS) then
print *, irp_here, 'error in barrier'
endif
IRP_ENDIF
IRP_IF MPI_DEBUG
call mpi_print('Entering OpenMP section')
IRP_ENDIF
if (.true.) then
integer :: nproc_target, ii
double precision :: mem_collector, mem, rss
call resident_memory(rss)
nproc_target = nthreads_pt2
ii = min(N_det, (elec_alpha_num*(mo_num-elec_alpha_num))**2)
do
mem = rss + & !
nproc_target * 8.d0 * & ! bytes
( 0.5d0*pt2_n_tasks_max & ! task_id
+ 64.d0*pt2_n_tasks_max & ! task
+ 3.d0*pt2_n_tasks_max*N_states & ! pt2, variance, norm
+ 1.d0*pt2_n_tasks_max & ! i_generator, subset
+ 3.d0*(N_int*2.d0*ii+ ii) & ! selection buffer
+ 1.d0*(N_int*2.d0*ii+ ii) & ! sort selection buffer
+ 2.0d0*(ii) & ! preinteresting, interesting,
! prefullinteresting, fullinteresting
+ 2.0d0*(N_int*2*ii) & ! minilist, fullminilist
+ 1.0d0*(N_states*mo_num*mo_num) & ! mat
) / 1024.d0**3
if (nproc_target == 0) then
call check_mem(mem,irp_here)
nproc_target = 1
exit
endif
if (mem+rss < qp_max_mem) then
exit
endif
nproc_target = nproc_target - 1
enddo
if (N_det > 100000) then
if (mpi_master) then
print *, 'N_det', N_det
print *, 'N_det_generators', N_det_generators
print *, 'N_det_selectors', N_det_selectors
print *, 'pt2_e0_denominator', pt2_e0_denominator
print *, 'pt2_stoch_istate', pt2_stoch_istate
print *, 'state_average_weight', state_average_weight
print *, 'selection_weight', selection_weight
print *, 'Number of threads', nproc_target
endif
if (h0_type == 'CFG') then
PROVIDE det_to_configuration
endif
PROVIDE global_selection_buffer pt2_N_teeth pt2_F N_det_generators
PROVIDE psi_bilinear_matrix_columns_loc psi_det_alpha_unique psi_det_beta_unique
PROVIDE psi_bilinear_matrix_rows psi_det_sorted_tc_order psi_bilinear_matrix_order
PROVIDE psi_bilinear_matrix_transp_rows_loc psi_bilinear_matrix_transp_columns
PROVIDE psi_bilinear_matrix_transp_order psi_selectors_coef_transp psi_det_sorted_tc
PROVIDE psi_selectors_rcoef_bi_orth_transp psi_selectors_lcoef_bi_orth_transp
PROVIDE psi_det_hii selection_weight pseudo_sym pt2_min_parallel_tasks
if (mpi_master) then
print *, 'Running PT2'
endif
!$OMP PARALLEL PRIVATE(i) NUM_THREADS(nproc_target+1)
i = omp_get_thread_num()
call run_pt2_slave(0,i,pt2_e0_denominator)
!$OMP END PARALLEL
FREE state_average_weight
print *, mpi_rank, ': PT2 done'
print *, '-------'
endif
endif
IRP_IF MPI
call MPI_BARRIER(MPI_COMM_WORLD, ierr)
if (ierr /= MPI_SUCCESS) then
print *, irp_here, 'error in barrier'
endif
IRP_ENDIF
call mpi_print('----------')
endif
end do
IRP_IF MPI
call MPI_finalize(ierr)
IRP_ENDIF
end

View File

@ -0,0 +1,147 @@
subroutine run_stochastic_cipsi
use selection_types
implicit none
BEGIN_DOC
! Selected Full Configuration Interaction with Stochastic selection and PT2.
END_DOC
integer :: i,j,k,ndet
double precision, allocatable :: zeros(:)
integer :: to_select
type(pt2_type) :: pt2_data, pt2_data_err
logical, external :: qp_stop
logical :: print_pt2
double precision :: rss
double precision, external :: memory_of_double
double precision :: correlation_energy_ratio,E_denom,E_tc,norm
double precision, allocatable :: ept2(:), pt1(:),extrap_energy(:)
PROVIDE H_apply_buffer_allocated distributed_davidson
print*,'Diagonal elements of the Fock matrix '
do i = 1, mo_num
write(*,*)i,Fock_matrix_tc_mo_tot(i,i)
enddo
N_iter = 1
threshold_generators = 1.d0
SOFT_TOUCH threshold_generators
rss = memory_of_double(N_states)*4.d0
call check_mem(rss,irp_here)
allocate (zeros(N_states))
call pt2_alloc(pt2_data, N_states)
call pt2_alloc(pt2_data_err, N_states)
double precision :: hf_energy_ref
logical :: has
double precision :: relative_error
relative_error=PT2_relative_error
zeros = 0.d0
pt2_data % pt2 = -huge(1.e0)
pt2_data % rpt2 = -huge(1.e0)
pt2_data % overlap= 0.d0
pt2_data % variance = huge(1.e0)
if (s2_eig) then
call make_s2_eigenfunction
endif
print_pt2 = .False.
call diagonalize_CI_tc_bi_ortho(ndet, E_tc,norm,pt2_data,print_pt2)
! call routine_save_right
if (N_det > N_det_max) then
psi_det(1:N_int,1:2,1:N_det) = psi_det_sorted_tc_gen(1:N_int,1:2,1:N_det)
psi_coef(1:N_det,1:N_states) = psi_coef_sorted_tc_gen(1:N_det,1:N_states)
N_det = N_det_max
soft_touch N_det psi_det psi_coef
if (s2_eig) then
call make_s2_eigenfunction
endif
print_pt2 = .False.
call diagonalize_CI_tc_bi_ortho(ndet, E_tc,norm,pt2_data,print_pt2)
! call routine_save_right
endif
allocate(ept2(1000),pt1(1000),extrap_energy(100))
correlation_energy_ratio = 0.d0
! thresh_it_dav = 5.d-5
! soft_touch thresh_it_dav
print_pt2 = .True.
do while ( &
(N_det < N_det_max) .and. &
(maxval(abs(pt2_data % pt2(1:N_states))) > pt2_max) &
)
write(*,'(A)') '--------------------------------------------------------------------------------'
to_select = int(sqrt(dble(N_states))*dble(N_det)*selection_factor)
to_select = max(N_states_diag, to_select)
E_denom = E_tc ! TC Energy of the current wave function
call pt2_dealloc(pt2_data)
call pt2_dealloc(pt2_data_err)
call pt2_alloc(pt2_data, N_states)
call pt2_alloc(pt2_data_err, N_states)
call ZMQ_pt2(E_denom, pt2_data, pt2_data_err, relative_error,to_select) ! Stochastic PT2 and selection
N_iter += 1
if (qp_stop()) exit
! Add selected determinants
call copy_H_apply_buffer_to_wf_tc()
PROVIDE psi_l_coef_bi_ortho psi_r_coef_bi_ortho
PROVIDE psi_det
PROVIDE psi_det_sorted_tc
ept2(N_iter-1) = E_tc + nuclear_repulsion + (pt2_data % pt2(1))/norm
pt1(N_iter-1) = dsqrt(pt2_data % overlap(1,1))
call diagonalize_CI_tc_bi_ortho(ndet, E_tc,norm,pt2_data,print_pt2)
if (qp_stop()) exit
enddo
! print*,'data to extrapolate '
! do i = 2, N_iter
! print*,'iteration ',i
! print*,'pt1,Ept2',pt1(i),ept2(i)
! call get_extrapolated_energy(i-1,ept2(i),pt1(i),extrap_energy(i))
! do j = 2, i
! print*,'j,e,energy',j,extrap_energy(j)
! enddo
! enddo
! thresh_it_dav = 5.d-6
! soft_touch thresh_it_dav
call pt2_dealloc(pt2_data)
call pt2_dealloc(pt2_data_err)
call pt2_alloc(pt2_data, N_states)
call pt2_alloc(pt2_data_err, N_states)
call ZMQ_pt2(E_tc, pt2_data, pt2_data_err, relative_error,0) ! Stochastic PT2 and selection
call diagonalize_CI_tc_bi_ortho(ndet, E_tc,norm,pt2_data,print_pt2)
! if (.not.qp_stop()) then
! if (N_det < N_det_max) then
! thresh_it_dav = 5.d-7
! soft_touch thresh_it_dav
! call diagonalize_CI_tc_bi_ortho(ndet, E_tc,norm,pt2_data,print_pt2)
! endif
!
! call pt2_dealloc(pt2_data)
! call pt2_dealloc(pt2_data_err)
! call pt2_alloc(pt2_data, N_states)
! call pt2_alloc(pt2_data_err, N_states)
! call ZMQ_pt2(E_denom, pt2_data, pt2_data_err, relative_error, 0) ! Stochastic PT2
! call diagonalize_CI_tc_bi_ortho(ndet, E_tc,norm,pt2_data,print_pt2)
! endif
! call pt2_dealloc(pt2_data)
! call pt2_dealloc(pt2_data_err)
! call routine_save_right
end

View File

@ -0,0 +1,235 @@
subroutine ZMQ_selection(N_in, pt2_data)
use f77_zmq
use selection_types
implicit none
integer(ZMQ_PTR) :: zmq_to_qp_run_socket , zmq_socket_pull
integer, intent(in) :: N_in
type(selection_buffer) :: b
integer :: i, l, N
integer, external :: omp_get_thread_num
type(pt2_type), intent(inout) :: pt2_data
PROVIDE psi_det psi_coef N_det qp_max_mem N_states pt2_F s2_eig N_det_generators
N = max(N_in,1)
N = min(N, (elec_alpha_num * (mo_num-elec_alpha_num))**2)
if (.True.) then
PROVIDE pt2_e0_denominator nproc
PROVIDE psi_bilinear_matrix_columns_loc psi_det_alpha_unique psi_det_beta_unique
PROVIDE psi_bilinear_matrix_rows psi_det_sorted_tc_order psi_bilinear_matrix_order
PROVIDE psi_bilinear_matrix_transp_rows_loc psi_bilinear_matrix_transp_columns
PROVIDE psi_bilinear_matrix_transp_order selection_weight pseudo_sym
PROVIDE n_act_orb n_inact_orb n_core_orb n_virt_orb n_del_orb seniority_max
PROVIDE excitation_beta_max excitation_alpha_max excitation_max
call new_parallel_job(zmq_to_qp_run_socket,zmq_socket_pull,'selection')
integer, external :: zmq_put_psi
integer, external :: zmq_put_N_det_generators
integer, external :: zmq_put_N_det_selectors
integer, external :: zmq_put_dvector
if (zmq_put_psi(zmq_to_qp_run_socket,1) == -1) then
stop 'Unable to put psi on ZMQ server'
endif
if (zmq_put_N_det_generators(zmq_to_qp_run_socket, 1) == -1) then
stop 'Unable to put N_det_generators on ZMQ server'
endif
if (zmq_put_N_det_selectors(zmq_to_qp_run_socket, 1) == -1) then
stop 'Unable to put N_det_selectors on ZMQ server'
endif
if (zmq_put_dvector(zmq_to_qp_run_socket,1,'energy',pt2_e0_denominator,size(pt2_e0_denominator)) == -1) then
stop 'Unable to put energy on ZMQ server'
endif
if (zmq_put_dvector(zmq_to_qp_run_socket,1,'state_average_weight',state_average_weight,N_states) == -1) then
stop 'Unable to put state_average_weight on ZMQ server'
endif
if (zmq_put_dvector(zmq_to_qp_run_socket,1,'selection_weight',selection_weight,N_states) == -1) then
stop 'Unable to put selection_weight on ZMQ server'
endif
if (zmq_put_dvector(zmq_to_qp_run_socket,1,'threshold_generators',(/threshold_generators/),1) == -1) then
stop 'Unable to put threshold_generators on ZMQ server'
endif
call create_selection_buffer(N, N*2, b)
endif
integer, external :: add_task_to_taskserver
character(len=100000) :: task
integer :: j,k,ipos
ipos=1
task = ' '
do i= 1, N_det_generators
do j=1,pt2_F(i)
write(task(ipos:ipos+30),'(I9,1X,I9,1X,I9,''|'')') j, i, N
ipos += 30
if (ipos > 100000-30) then
if (add_task_to_taskserver(zmq_to_qp_run_socket,trim(task(1:ipos))) == -1) then
stop 'Unable to add task to task server'
endif
ipos=1
endif
end do
enddo
if (ipos > 1) then
if (add_task_to_taskserver(zmq_to_qp_run_socket,trim(task(1:ipos))) == -1) then
stop 'Unable to add task to task server'
endif
endif
N = max(N_in,1)
ASSERT (associated(b%det))
ASSERT (associated(b%val))
integer, external :: zmq_set_running
if (zmq_set_running(zmq_to_qp_run_socket) == -1) then
print *, irp_here, ': Failed in zmq_set_running'
endif
integer :: nproc_target
if (N_det < 3*nproc) then
nproc_target = N_det/4
else
nproc_target = nproc
endif
double precision :: mem
mem = 8.d0 * N_det * (N_int * 2.d0 * 3.d0 + 3.d0 + 5.d0) / (1024.d0**3)
call write_double(6,mem,'Estimated memory/thread (Gb)')
if (qp_max_mem > 0) then
nproc_target = max(1,int(dble(qp_max_mem)/(0.1d0 + mem)))
nproc_target = min(nproc_target,nproc)
endif
f(:) = 1.d0
if (.not.do_pt2) then
double precision :: f(N_states), u_dot_u
do k=1,min(N_det,N_states)
f(k) = 1.d0 / u_dot_u(psi_selectors_coef(1,k), N_det_selectors)
enddo
endif
!$OMP PARALLEL DEFAULT(shared) SHARED(b, pt2_data) PRIVATE(i) NUM_THREADS(nproc_target+1)
i = omp_get_thread_num()
if (i==0) then
call selection_collector(zmq_socket_pull, b, N, pt2_data)
else
call selection_slave_inproc(i)
endif
!$OMP END PARALLEL
call end_parallel_job(zmq_to_qp_run_socket, zmq_socket_pull, 'selection')
if (N_in > 0) then
if (s2_eig) then
call make_selection_buffer_s2(b)
endif
call fill_H_apply_buffer_no_selection(b%cur,b%det,N_int,0)
endif
call delete_selection_buffer(b)
do k=1,N_states
pt2_data % pt2(k) = pt2_data % pt2(k) * f(k)
pt2_data % variance(k) = pt2_data % variance(k) * f(k)
do l=1,N_states
pt2_data % overlap(k,l) = pt2_data % overlap(k,l) * dsqrt(f(k)*f(l))
pt2_data % overlap(l,k) = pt2_data % overlap(l,k) * dsqrt(f(k)*f(l))
enddo
pt2_data % rpt2(k) = &
pt2_data % pt2(k)/(1.d0 + pt2_data % overlap(k,k))
enddo
pt2_overlap(:,:) = pt2_data % overlap(:,:)
print *, 'Overlap of perturbed states:'
do l=1,N_states
print *, pt2_overlap(l,:)
enddo
print *, '-------'
SOFT_TOUCH pt2_overlap
call update_pt2_and_variance_weights(pt2_data, N_states)
end subroutine
subroutine selection_slave_inproc(i)
implicit none
integer, intent(in) :: i
call run_selection_slave(1,i,pt2_e0_denominator)
end
subroutine selection_collector(zmq_socket_pull, b, N, pt2_data)
use f77_zmq
use selection_types
use bitmasks
implicit none
integer(ZMQ_PTR), intent(in) :: zmq_socket_pull
type(selection_buffer), intent(inout) :: b
integer, intent(in) :: N
type(pt2_type), intent(inout) :: pt2_data
type(pt2_type) :: pt2_data_tmp
double precision :: pt2_mwen(N_states)
double precision :: variance_mwen(N_states)
double precision :: norm2_mwen(N_states)
integer(ZMQ_PTR),external :: new_zmq_to_qp_run_socket
integer(ZMQ_PTR) :: zmq_to_qp_run_socket
integer(ZMQ_PTR), external :: new_zmq_pull_socket
integer :: msg_size, rc, more
integer :: acc, i, j, robin, ntask
double precision, pointer :: val(:)
integer(bit_kind), pointer :: det(:,:,:)
integer, allocatable :: task_id(:)
type(selection_buffer) :: b2
zmq_to_qp_run_socket = new_zmq_to_qp_run_socket()
call create_selection_buffer(N, N*2, b2)
integer :: k
double precision :: rss
double precision, external :: memory_of_int
rss = memory_of_int(N_det_generators)
call check_mem(rss,irp_here)
allocate(task_id(N_det_generators))
more = 1
pt2_data % pt2(:) = 0d0
pt2_data % variance(:) = 0.d0
pt2_data % overlap(:,:) = 0.d0
call pt2_alloc(pt2_data_tmp,N_states)
do while (more == 1)
call pull_selection_results(zmq_socket_pull, pt2_data_tmp, b2%val(1), b2%det(1,1,1), b2%cur, task_id, ntask)
call pt2_add(pt2_data, 1.d0, pt2_data_tmp)
do i=1, b2%cur
call add_to_selection_buffer(b, b2%det(1,1,i), b2%val(i))
if (b2%val(i) > b%mini) exit
end do
do i=1, ntask
if(task_id(i) == 0) then
print *, "Error in collector"
endif
integer, external :: zmq_delete_task
if (zmq_delete_task(zmq_to_qp_run_socket,zmq_socket_pull,task_id(i),more) == -1) then
stop 'Unable to delete task'
endif
end do
end do
call pt2_dealloc(pt2_data_tmp)
call delete_selection_buffer(b2)
call sort_selection_buffer(b)
call end_zmq_to_qp_run_socket(zmq_to_qp_run_socket)
end subroutine