LCPQ/quantum_package
10
0
Fork 0
mirror of https://github.com/LCPQ/quantum_package synced 2024-07-31 17:44:23 +02:00
Code Issues Releases Wiki Activity
bddd875af7
quantum_package/src/Davidson/u0Hu0.irp.f

1037 lines
31 KiB
Fortran
Raw Normal View History

Optimized ninja
2016-09-23 10:01:26 +02:00
subroutine u_0_H_u_0(e_0,u_0,n,keys_tmp,Nint,N_st,sze_8)
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
use bitmasks
implicit none
BEGIN_DOC
! Computes e_0 = <u_0|H|u_0>/<u_0|u_0>
!
! n : number of determinants
!
END_DOC
integer, intent(in) :: n,Nint, N_st, sze_8
double precision, intent(out) :: e_0(N_st)
double precision, intent(in) :: u_0(sze_8,N_st)
integer(bit_kind),intent(in) :: keys_tmp(Nint,2,n)
double precision, allocatable :: H_jj(:), v_0(:,:)
double precision :: u_dot_u,u_dot_v,diag_H_mat_elem
integer :: i,j
allocate (H_jj(n), v_0(sze_8,N_st))
do i = 1, n
H_jj(i) = diag_H_mat_elem(keys_tmp(1,1,i),Nint)
enddo
call H_u_0_nstates(v_0,u_0,H_jj,n,keys_tmp,Nint,N_st,sze_8)
do i=1,N_st
e_0(i) = u_dot_v(v_0(1,i),u_0(1,i),n)/u_dot_u(u_0(1,i),n)
enddo
Reduced memory
2016-10-09 22:51:40 +02:00
deallocate (H_jj, v_0)
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
end
subroutine H_u_0_nstates(v_0,u_0,H_jj,n,keys_tmp,Nint,N_st,sze_8)
use bitmasks
implicit none
BEGIN_DOC
MRCC_selected
2016-12-29 22:00:41 +01:00
! Computes v_0 = H|u_0>
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
!
! n : number of determinants
!
! H_jj : array of <j|H|j>
MRCC_selected
2016-12-29 22:00:41 +01:00
!
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
END_DOC
integer, intent(in) :: N_st,n,Nint, sze_8
double precision, intent(out) :: v_0(sze_8,N_st)
double precision, intent(in) :: u_0(sze_8,N_st)
double precision, intent(in) :: H_jj(n)
integer(bit_kind),intent(in) :: keys_tmp(Nint,2,n)
MRCC_selected
2016-12-29 22:00:41 +01:00
double precision :: hij,s2
double precision, allocatable :: vt(:,:), ut(:,:), st(:,:)
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
integer :: i,j,k,l, jj,ii
integer :: i0, j0
integer, allocatable :: shortcut(:,:), sort_idx(:,:)
integer(bit_kind), allocatable :: sorted(:,:,:), version(:,:,:)
integer(bit_kind) :: sorted_i(Nint)
integer :: sh, sh2, ni, exa, ext, org_i, org_j, endi, istate
Working on davidson
2016-09-23 16:16:48 +02:00
integer :: N_st_8
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
Working on davidson
2016-09-23 16:16:48 +02:00
integer, external :: align_double
MRCC_selected
2016-12-29 22:00:41 +01:00
!DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: vt, ut, st
Working on davidson
2016-09-23 16:16:48 +02:00
Fixed bug of dimensions in davidson
2016-10-07 12:22:43 +02:00
N_st_8 = align_double(N_st)
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
ASSERT (Nint > 0)
ASSERT (Nint == N_int)
ASSERT (n>0)
MRCC_selected
2016-12-29 22:00:41 +01:00
PROVIDE ref_bitmask_energy
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
allocate (shortcut(0:n+1,2), sort_idx(n,2), sorted(Nint,n,2), version(Nint,n,2))
MRCC_selected
2016-12-29 22:00:41 +01:00
allocate( ut(N_st_8,n))
Working on davidson
2016-09-23 16:16:48 +02:00
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
v_0 = 0.d0
MRCC_selected
2016-12-29 22:00:41 +01:00
call sort_dets_ab_v(keys_tmp, sorted(1,1,1), sort_idx(1,1), shortcut(0,1), version(1,1,1), n, Nint)
call sort_dets_ba_v(keys_tmp, sorted(1,1,2), sort_idx(1,2), shortcut(0,2), version(1,1,2), n, Nint)
!$OMP PARALLEL DEFAULT(NONE) &
!$OMP PRIVATE(i,hij,s2,j,k,jj,vt,st,ii,sh,sh2,ni,exa,ext,org_i,org_j,endi,sorted_i,istate)&
!$OMP SHARED(n,keys_tmp,ut,Nint,u_0,v_0,sorted,shortcut,sort_idx,version,N_st,N_st_8)
allocate(vt(N_st_8,n),st(N_st_8,n))
Vt = 0.d0
St = 0.d0
!$OMP DO
Working on davidson
2016-09-23 16:16:48 +02:00
do i=1,n
do istate=1,N_st
MRCC_selected
2016-12-29 22:00:41 +01:00
ut(istate,i) = u_0(sort_idx(i,2),istate)
Working on davidson
2016-09-23 16:16:48 +02:00
enddo
enddo
MRCC_selected
2016-12-29 22:00:41 +01:00
!$OMP END DO
openmp
2017-04-12 15:39:16 +02:00
!$OMP DO SCHEDULE(static,1)
MRCC_selected
2016-12-29 22:00:41 +01:00
do sh=1,shortcut(0,2)
do i=shortcut(sh,2),shortcut(sh+1,2)-1
org_i = sort_idx(i,2)
do j=shortcut(sh,2),shortcut(sh+1,2)-1
org_j = sort_idx(j,2)
ext = popcnt(xor(sorted(1,i,2), sorted(1,j,2)))
if (ext > 4) cycle
do ni=2,Nint
ext = ext + popcnt(xor(sorted(ni,i,2), sorted(ni,j,2)))
if (ext > 4) exit
end do
if(ext == 4) then
call i_h_j (keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,hij)
call get_s2(keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,s2)
do istate=1,n_st
vt (istate,org_i) = vt (istate,org_i) + hij*ut(istate,j)
st (istate,org_i) = st (istate,org_i) + s2*ut(istate,j)
enddo
end if
end do
end do
enddo
!$OMP END DO
!$OMP DO
do i=1,n
do istate=1,N_st
ut(istate,i) = u_0(sort_idx(i,1),istate)
enddo
enddo
!$OMP END DO
Working on davidson
2016-09-23 16:16:48 +02:00
openmp
2017-04-12 15:39:16 +02:00
!$OMP DO SCHEDULE(static,1)
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
do sh=1,shortcut(0,1)
De-symmetrized u0_H_u0
2016-12-13 12:29:48 +01:00
do sh2=1,shortcut(0,1)
MRCC_selected
2016-12-29 22:00:41 +01:00
if (sh==sh2) cycle
exa = 0
do ni=1,Nint
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
exa = exa + popcnt(xor(version(ni,sh,1), version(ni,sh2,1)))
end do
if(exa > 2) then
cycle
end if
MRCC_selected
2016-12-29 22:00:41 +01:00
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
do i=shortcut(sh,1),shortcut(sh+1,1)-1
org_i = sort_idx(i,1)
do ni=1,Nint
sorted_i(ni) = sorted(ni,i,1)
enddo
MRCC_selected
2016-12-29 22:00:41 +01:00
do j=shortcut(sh2,1),shortcut(sh2+1,1)-1
Acceleration of Davidson for Nint>1
2016-12-06 11:31:15 +01:00
ext = exa + popcnt(xor(sorted_i(1), sorted(1,j,1)))
MRCC_selected
2016-12-29 22:00:41 +01:00
if (ext > 4) cycle
Acceleration of Davidson for Nint>1
2016-12-06 11:31:15 +01:00
do ni=2,Nint
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
ext = ext + popcnt(xor(sorted_i(ni), sorted(ni,j,1)))
MRCC_selected
2016-12-29 22:00:41 +01:00
if (ext > 4) exit
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
end do
MRCC_selected
2016-12-29 22:00:41 +01:00
if(ext <= 4) then
org_j = sort_idx(j,1)
call i_h_j (keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,hij)
if (hij /= 0.d0) then
do istate=1,n_st
vt (istate,org_i) = vt (istate,org_i) + hij*ut(istate,j)
enddo
endif
if (ext /= 2) then
call get_s2(keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,s2)
if (s2 /= 0.d0) then
do istate=1,n_st
st (istate,org_i) = st (istate,org_i) + s2*ut(istate,j)
enddo
endif
endif
endif
enddo
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
enddo
enddo
MRCC_selected
2016-12-29 22:00:41 +01:00
exa = 0
do i=shortcut(sh,1),shortcut(sh+1,1)-1
org_i = sort_idx(i,1)
do ni=1,Nint
sorted_i(ni) = sorted(ni,i,1)
enddo
do j=shortcut(sh,1),i-1
ext = exa + popcnt(xor(sorted_i(1), sorted(1,j,1)))
if (ext > 4) cycle
Acceleration of Davidson for Nint>1
2016-12-06 11:31:15 +01:00
do ni=2,Nint
MRCC_selected
2016-12-29 22:00:41 +01:00
ext = ext + popcnt(xor(sorted_i(ni), sorted(ni,j,1)))
if (ext > 4) exit
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
end do
MRCC_selected
2016-12-29 22:00:41 +01:00
if(ext <= 4) then
org_j = sort_idx(j,1)
call i_h_j (keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,hij)
if (hij /= 0.d0) then
do istate=1,n_st
vt (istate,org_i) = vt (istate,org_i) + hij*ut(istate,j)
enddo
endif
if (ext /= 2) then
call get_s2(keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,s2)
if (s2 /= 0.d0) then
do istate=1,n_st
st (istate,org_i) = st (istate,org_i) + s2*ut(istate,j)
enddo
endif
endif
Acceleration of Davidson for Nint>1
2016-12-06 11:31:15 +01:00
endif
MRCC_selected
2016-12-29 22:00:41 +01:00
enddo
do j=i+1,shortcut(sh+1,1)-1
if (i==j) cycle
ext = exa + popcnt(xor(sorted_i(1), sorted(1,j,1)))
if (ext > 4) cycle
do ni=2,Nint
ext = ext + popcnt(xor(sorted_i(ni), sorted(ni,j,1)))
if (ext > 4) exit
end do
if(ext <= 4) then
org_j = sort_idx(j,1)
call i_h_j (keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,hij)
if (hij /= 0.d0) then
do istate=1,n_st
vt (istate,org_i) = vt (istate,org_i) + hij*ut(istate,j)
enddo
endif
if (ext /= 2) then
call get_s2(keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,s2)
if (s2 /= 0.d0) then
do istate=1,n_st
st (istate,org_i) = st (istate,org_i) + s2*ut(istate,j)
enddo
endif
endif
endif
enddo
enddo
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
enddo
MRCC_selected
2016-12-29 22:00:41 +01:00
!$OMP END DO
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
do istate=1,N_st
MRCC_selected
2016-12-29 22:00:41 +01:00
do i=1,n
Optims davidson
2017-03-06 18:29:39 +01:00
!$OMP ATOMIC
Working on davidson
2016-09-23 16:16:48 +02:00
v_0(i,istate) = v_0(i,istate) + vt(istate,i)
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
enddo
enddo
MRCC_selected
2016-12-29 22:00:41 +01:00
deallocate(vt,st)
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
!$OMP END PARALLEL
MRCC_selected
2016-12-29 22:00:41 +01:00
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
do istate=1,N_st
do i=1,n
MRCC_selected
2016-12-29 22:00:41 +01:00
v_0(i,istate) = v_0(i,istate) + H_jj(i) * u_0(i,istate)
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
enddo
enddo
Fixed bug of dimensions in davidson
2016-10-07 12:22:43 +02:00
deallocate (shortcut, sort_idx, sorted, version, ut)
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
end
MRCC_selected
2016-12-29 22:00:41 +01:00
moved psi_energy
2016-09-22 16:26:54 +02:00
BEGIN_PROVIDER [ double precision, psi_energy, (N_states) ]
implicit none
BEGIN_DOC
! Energy of the current wave function
END_DOC
Optimized ninja
2016-09-23 10:01:26 +02:00
call u_0_H_u_0(psi_energy,psi_coef,N_det,psi_det,N_int,N_states,psi_det_size)
moved psi_energy
2016-09-22 16:26:54 +02:00
END_PROVIDER
Extracting davidson from slater rules
2016-09-20 17:29:02 +02:00
Include S^2 inside davidson
2016-09-27 15:55:38 +02:00
Accelerated distributed Davidson
2017-03-01 01:19:17 +01:00
subroutine H_S2_u_0_nstates_zmq(v_0,s_0,u_0,H_jj,S2_jj,n,keys_tmp,Nint,N_st,sze_8,update_dets)
Fixed distributed Davidson
2017-02-08 20:54:15 +01:00
use omp_lib
Include S^2 inside davidson
2016-09-27 15:55:38 +02:00
use bitmasks
shared memory davidson
2016-09-30 15:29:06 +02:00
use f77_zmq
Include S^2 inside davidson
2016-09-27 15:55:38 +02:00
implicit none
BEGIN_DOC
! Computes v_0 = H|u_0> and s_0 = S^2 |u_0>
!
! n : number of determinants
!
! H_jj : array of <j|H|j>
!
! S2_jj : array of <j|S^2|j>
END_DOC
Accelerated distributed Davidson
2017-03-01 01:19:17 +01:00
integer, intent(in) :: N_st,n,Nint, sze_8, update_dets
Include S^2 inside davidson
2016-09-27 15:55:38 +02:00
double precision, intent(out) :: v_0(sze_8,N_st), s_0(sze_8,N_st)
double precision, intent(in) :: u_0(sze_8,N_st)
double precision, intent(in) :: H_jj(n), S2_jj(n)
integer(bit_kind),intent(in) :: keys_tmp(Nint,2,n)
double precision :: hij,s2
integer :: i,j,k,l, jj,ii
Fixed distributed Davidson
2017-02-08 20:54:15 +01:00
integer :: i0, j0, ithread
Include S^2 inside davidson
2016-09-27 15:55:38 +02:00
integer(bit_kind) :: sorted_i(Nint)
integer :: sh, sh2, ni, exa, ext, org_i, org_j, endi, istate
integer :: N_st_8
integer, external :: align_double
Memory in Davidson
2017-02-27 10:43:57 +01:00
integer :: blockb2, istep
Manu changes
2016-11-04 17:34:05 +01:00
double precision :: ave_workload, workload, target_workload_inv
shared memory davidson
2016-09-30 15:29:06 +02:00
integer(ZMQ_PTR) :: handler
Corrected N_states N_states_diag i parallel davidson
2016-10-06 14:35:43 +02:00
if(N_st /= N_states_diag .or. sze_8 < N_det) stop "assert fail in H_S2_u_0_nstates"
Fixed bug of dimensions in davidson
2016-10-07 12:22:43 +02:00
N_st_8 = N_st ! align_double(N_st)
Include S^2 inside davidson
2016-09-27 15:55:38 +02:00
ASSERT (Nint > 0)
ASSERT (Nint == N_int)
ASSERT (n>0)
PROVIDE ref_bitmask_energy
v_0 = 0.d0
s_0 = 0.d0
hobo_server - unstable
2016-09-30 18:33:46 +02:00
ZMQ checks
2017-03-25 11:57:06 +01:00
call davidson_init(handler,keys_tmp,u_0,size(u_0,1),n,N_st,update_dets)
Memory in Davidson
2017-02-27 10:43:57 +01:00
ave_workload = 0.d0
do sh=1,shortcut_(0,1)
ave_workload += shortcut_(0,1)
ave_workload += (shortcut_(sh+1,1) - shortcut_(sh,1))**2
do i=sh, shortcut_(0,2), shortcut_(0,1)
do j=i, min(i, shortcut_(0,2))
ave_workload += (shortcut_(j+1,2) - shortcut_(j, 2))**2
end do
end do
Include S^2 inside davidson
2016-09-27 15:55:38 +02:00
enddo
Memory in Davidson
2017-02-27 10:43:57 +01:00
ave_workload = ave_workload/dble(shortcut_(0,1))
target_workload_inv = 0.01d0/ave_workload
Improved parallel davidson
2016-10-11 22:45:40 +02:00
Fixed distributed Davidson
2017-02-08 20:54:15 +01:00
PROVIDE nproc
Memory in Davidson
2017-02-27 10:43:57 +01:00
Less pressure on qp_run when ading tasks
2017-02-28 22:43:59 +01:00
Fixed communications
2017-03-03 22:20:57 +01:00
character(len=:), allocatable :: task
task = repeat(' ', iposmax)
character(32) :: tmp_task
integer :: ipos, iposmax
iposmax = shortcut_(0,1)+32
ipos = 1
do sh=1,shortcut_(0,1),1
workload = shortcut_(0,1)+dble(shortcut_(sh+1,1) - shortcut_(sh,1))**2
do i=sh, shortcut_(0,2), shortcut_(0,1)
do j=i, min(i, shortcut_(0,2))
workload += (shortcut_(j+1,2) - shortcut_(j, 2))**2
Improved parallel davidson
2016-10-11 22:45:40 +02:00
end do
Fixed communications
2017-03-03 22:20:57 +01:00
end do
Temporary fix for parallel davdison
2017-03-25 11:56:08 +01:00
! istep = 1+ int(workload*target_workload_inv)
istep = 1
Fixed communications
2017-03-03 22:20:57 +01:00
do blockb2=0, istep-1
pedantic changes
2017-04-12 19:29:21 +02:00
write(tmp_task,'(3(I9,1X),''|'',1X)') sh, blockb2, istep
Fixed communications
2017-03-03 22:20:57 +01:00
task = task//tmp_task
ipos += 32
if (ipos+32 > iposmax) then
call add_task_to_taskserver(handler, trim(task))
ipos=1
task = ''
endif
Improved parallel davidson
2016-10-11 22:45:40 +02:00
enddo
Fixed communications
2017-03-03 22:20:57 +01:00
enddo
if (ipos>1) then
call add_task_to_taskserver(handler, trim(task))
endif
!$OMP PARALLEL NUM_THREADS(nproc+2) PRIVATE(ithread)
ithread = omp_get_thread_num()
if (ithread == 0 ) then
Memory in Davidson
2017-02-27 10:43:57 +01:00
call zmq_set_running(handler)
Fixed distributed Davidson
2017-02-08 20:54:15 +01:00
call davidson_run(handler, v_0, s_0, size(v_0,1))
else if (ithread == 1 ) then
Accelerated distributed Davidson
2017-03-01 01:19:17 +01:00
call davidson_miniserver_run (update_dets)
Fixed distributed Davidson
2017-02-08 20:54:15 +01:00
else
call davidson_slave_inproc(ithread)
endif
!$OMP END PARALLEL
call end_parallel_job(handler, 'davidson')
Improve parallel davidson
2016-10-07 00:42:03 +02:00
Include S^2 inside davidson
2016-09-27 15:55:38 +02:00
do istate=1,N_st
do i=1,n
v_0(i,istate) = v_0(i,istate) + H_jj(i) * u_0(i,istate)
s_0(i,istate) = s_0(i,istate) + s2_jj(i)* u_0(i,istate)
enddo
enddo
end
removed OMP - double work
2016-10-10 14:45:58 +02:00
Removed ZMQ parallelization in Davidson
2016-11-08 16:59:29 +01:00
subroutine H_S2_u_0_nstates(v_0,s_0,u_0,H_jj,S2_jj,n,keys_tmp,Nint,N_st,sze_8)
use bitmasks
implicit none
BEGIN_DOC
! Computes v_0 = H|u_0> and s_0 = S^2 |u_0>
!
! n : number of determinants
!
! H_jj : array of <j|H|j>
!
! S2_jj : array of <j|S^2|j>
END_DOC
integer, intent(in) :: N_st,n,Nint, sze_8
double precision, intent(out) :: v_0(sze_8,N_st), s_0(sze_8,N_st)
double precision, intent(in) :: u_0(sze_8,N_st)
double precision, intent(in) :: H_jj(n), S2_jj(n)
integer(bit_kind),intent(in) :: keys_tmp(Nint,2,n)
double precision :: hij,s2
double precision, allocatable :: vt(:,:), ut(:,:), st(:,:)
integer :: i,j,k,l, jj,ii
integer :: i0, j0
integer, allocatable :: shortcut(:,:), sort_idx(:,:)
integer(bit_kind), allocatable :: sorted(:,:,:), version(:,:,:)
integer(bit_kind) :: sorted_i(Nint)
integer :: sh, sh2, ni, exa, ext, org_i, org_j, endi, istate
integer :: N_st_8
integer, external :: align_double
Minor changes
2016-12-03 18:58:07 +01:00
!DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: vt, ut, st
Removed ZMQ parallelization in Davidson
2016-11-08 16:59:29 +01:00
N_st_8 = align_double(N_st)
ASSERT (Nint > 0)
ASSERT (Nint == N_int)
ASSERT (n>0)
PROVIDE ref_bitmask_energy
allocate (shortcut(0:n+1,2), sort_idx(n,2), sorted(Nint,n,2), version(Nint,n,2))
Fixed many bugs
2017-02-10 03:24:12 +01:00
allocate( ut(N_st_8,n))
Removed ZMQ parallelization in Davidson
2016-11-08 16:59:29 +01:00
v_0 = 0.d0
s_0 = 0.d0
call sort_dets_ab_v(keys_tmp, sorted(1,1,1), sort_idx(1,1), shortcut(0,1), version(1,1,1), n, Nint)
call sort_dets_ba_v(keys_tmp, sorted(1,1,2), sort_idx(1,2), shortcut(0,2), version(1,1,2), n, Nint)
!$OMP PARALLEL DEFAULT(NONE) &
!$OMP PRIVATE(i,hij,s2,j,k,jj,vt,st,ii,sh,sh2,ni,exa,ext,org_i,org_j,endi,sorted_i,istate)&
Fixed many bugs
2017-02-10 03:24:12 +01:00
!$OMP SHARED(n,keys_tmp,ut,Nint,u_0,v_0,s_0,sorted,shortcut,sort_idx,version,N_st,N_st_8)
Removed ZMQ parallelization in Davidson
2016-11-08 16:59:29 +01:00
allocate(vt(N_st_8,n),st(N_st_8,n))
Vt = 0.d0
St = 0.d0
De-symmetrized u0_H_u0
2016-12-13 12:29:48 +01:00
!$OMP DO
do i=1,n
do istate=1,N_st
ut(istate,i) = u_0(sort_idx(i,2),istate)
enddo
enddo
!$OMP END DO
Accelerated selection
2017-03-30 01:13:27 +02:00
!$OMP DO SCHEDULE(static,4)
Better parallelization of Davidson
2016-11-14 13:33:36 +01:00
do sh=1,shortcut(0,2)
do i=shortcut(sh,2),shortcut(sh+1,2)-1
org_i = sort_idx(i,2)
De-symmetrized u0_H_u0
2016-12-13 12:29:48 +01:00
do j=shortcut(sh,2),shortcut(sh+1,2)-1
Better parallelization of Davidson
2016-11-14 13:33:36 +01:00
org_j = sort_idx(j,2)
Acceleration of Davidson for Nint>1
2016-12-06 11:31:15 +01:00
ext = popcnt(xor(sorted(1,i,2), sorted(1,j,2)))
if (ext > 4) cycle
do ni=2,Nint
Better parallelization of Davidson
2016-11-14 13:33:36 +01:00
ext = ext + popcnt(xor(sorted(ni,i,2), sorted(ni,j,2)))
Minor changes
2016-12-03 18:58:07 +01:00
if (ext > 4) exit
Better parallelization of Davidson
2016-11-14 13:33:36 +01:00
end do
if(ext == 4) then
call i_h_j (keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,hij)
call get_s2(keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,s2)
do istate=1,n_st
De-symmetrized u0_H_u0
2016-12-13 12:29:48 +01:00
vt (istate,org_i) = vt (istate,org_i) + hij*ut(istate,j)
st (istate,org_i) = st (istate,org_i) + s2*ut(istate,j)
Better parallelization of Davidson
2016-11-14 13:33:36 +01:00
enddo
end if
end do
end do
enddo
De-symmetrized u0_H_u0
2016-12-13 12:29:48 +01:00
!$OMP END DO
!$OMP DO
do i=1,n
do istate=1,N_st
ut(istate,i) = u_0(sort_idx(i,1),istate)
enddo
enddo
!$OMP END DO
Accelerated selection
2017-03-30 01:13:27 +02:00
!$OMP DO SCHEDULE(static,4)
Removed ZMQ parallelization in Davidson
2016-11-08 16:59:29 +01:00
do sh=1,shortcut(0,1)
De-symmetrized u0_H_u0
2016-12-13 12:29:48 +01:00
do sh2=1,shortcut(0,1)
Accelerated u0_h_u0
2016-12-13 12:40:12 +01:00
if (sh==sh2) cycle
Removed ZMQ parallelization in Davidson
2016-11-08 16:59:29 +01:00
exa = 0
do ni=1,Nint
exa = exa + popcnt(xor(version(ni,sh,1), version(ni,sh2,1)))
end do
if(exa > 2) then
cycle
end if
do i=shortcut(sh,1),shortcut(sh+1,1)-1
org_i = sort_idx(i,1)
do ni=1,Nint
sorted_i(ni) = sorted(ni,i,1)
enddo
De-symmetrized u0_H_u0
2016-12-13 12:29:48 +01:00
do j=shortcut(sh2,1),shortcut(sh2+1,1)-1
Acceleration of Davidson for Nint>1
2016-12-06 11:31:15 +01:00
ext = exa + popcnt(xor(sorted_i(1), sorted(1,j,1)))
if (ext > 4) cycle
do ni=2,Nint
Removed ZMQ parallelization in Davidson
2016-11-08 16:59:29 +01:00
ext = ext + popcnt(xor(sorted_i(ni), sorted(ni,j,1)))
Minor changes
2016-12-03 18:58:07 +01:00
if (ext > 4) exit
Removed ZMQ parallelization in Davidson
2016-11-08 16:59:29 +01:00
end do
if(ext <= 4) then
Better parallelization of Davidson
2016-11-14 13:33:36 +01:00
org_j = sort_idx(j,1)
Removed ZMQ parallelization in Davidson
2016-11-08 16:59:29 +01:00
call i_h_j (keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,hij)
Better parallelization of Davidson
2016-11-14 13:33:36 +01:00
if (hij /= 0.d0) then
do istate=1,n_st
De-symmetrized u0_H_u0
2016-12-13 12:29:48 +01:00
vt (istate,org_i) = vt (istate,org_i) + hij*ut(istate,j)
Better parallelization of Davidson
2016-11-14 13:33:36 +01:00
enddo
endif
if (ext /= 2) then
call get_s2(keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,s2)
if (s2 /= 0.d0) then
do istate=1,n_st
De-symmetrized u0_H_u0
2016-12-13 12:29:48 +01:00
st (istate,org_i) = st (istate,org_i) + s2*ut(istate,j)
Better parallelization of Davidson
2016-11-14 13:33:36 +01:00
enddo
endif
endif
Removed ZMQ parallelization in Davidson
2016-11-08 16:59:29 +01:00
endif
enddo
Accelerated u0_h_u0
2016-12-13 12:40:12 +01:00
enddo
enddo
exa = 0
do i=shortcut(sh,1),shortcut(sh+1,1)-1
org_i = sort_idx(i,1)
do ni=1,Nint
sorted_i(ni) = sorted(ni,i,1)
enddo
do j=shortcut(sh,1),i-1
ext = exa + popcnt(xor(sorted_i(1), sorted(1,j,1)))
if (ext > 4) cycle
do ni=2,Nint
ext = ext + popcnt(xor(sorted_i(ni), sorted(ni,j,1)))
if (ext > 4) exit
end do
if(ext <= 4) then
org_j = sort_idx(j,1)
call i_h_j (keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,hij)
if (hij /= 0.d0) then
do istate=1,n_st
vt (istate,org_i) = vt (istate,org_i) + hij*ut(istate,j)
enddo
endif
if (ext /= 2) then
call get_s2(keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,s2)
if (s2 /= 0.d0) then
do istate=1,n_st
st (istate,org_i) = st (istate,org_i) + s2*ut(istate,j)
enddo
endif
endif
endif
enddo
do j=i+1,shortcut(sh+1,1)-1
ext = exa + popcnt(xor(sorted_i(1), sorted(1,j,1)))
if (ext > 4) cycle
do ni=2,Nint
ext = ext + popcnt(xor(sorted_i(ni), sorted(ni,j,1)))
if (ext > 4) exit
end do
if(ext <= 4) then
org_j = sort_idx(j,1)
call i_h_j (keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,hij)
if (hij /= 0.d0) then
do istate=1,n_st
vt (istate,org_i) = vt (istate,org_i) + hij*ut(istate,j)
enddo
endif
if (ext /= 2) then
call get_s2(keys_tmp(1,1,org_j),keys_tmp(1,1,org_i),nint,s2)
if (s2 /= 0.d0) then
do istate=1,n_st
st (istate,org_i) = st (istate,org_i) + s2*ut(istate,j)
enddo
endif
endif
endif
Removed ZMQ parallelization in Davidson
2016-11-08 16:59:29 +01:00
enddo
enddo
enddo
De-symmetrized u0_H_u0
2016-12-13 12:29:48 +01:00
!$OMP END DO
Removed ZMQ parallelization in Davidson
2016-11-08 16:59:29 +01:00
do istate=1,N_st
Introduced davidson_diag_hjj_sjj_mmap
2016-11-16 10:17:37 +01:00
do i=1,n
Optims davidson
2017-03-06 18:29:39 +01:00
!$OMP ATOMIC
Removed ZMQ parallelization in Davidson
2016-11-08 16:59:29 +01:00
v_0(i,istate) = v_0(i,istate) + vt(istate,i)
Optims davidson
2017-03-06 18:29:39 +01:00
!$OMP ATOMIC
Removed ZMQ parallelization in Davidson
2016-11-08 16:59:29 +01:00
s_0(i,istate) = s_0(i,istate) + st(istate,i)
enddo
enddo
deallocate(vt,st)
!$OMP END PARALLEL
do istate=1,N_st
do i=1,n
v_0(i,istate) = v_0(i,istate) + H_jj(i) * u_0(i,istate)
s_0(i,istate) = s_0(i,istate) + s2_jj(i)* u_0(i,istate)
enddo
enddo
deallocate (shortcut, sort_idx, sorted, version, ut)
end
Debug in davidson
2017-04-01 00:14:09 +02:00
New Davidson OK
2017-04-14 15:04:29 +02:00
subroutine H_S2_u_0_nstates_openmp(v_0,s_0,u_0,N_st,sze_8)
Debug in davidson
2017-04-01 00:14:09 +02:00
use bitmasks
implicit none
BEGIN_DOC
! Computes v_0 = H|u_0> and s_0 = S^2 |u_0>
!
New Davidson OK
2017-04-14 15:04:29 +02:00
! Assumes that the determinants are in psi_det
Debug in davidson
2017-04-01 00:14:09 +02:00
END_DOC
integer, intent(in) :: N_st,sze_8
New Davidson OK
2017-04-14 15:04:29 +02:00
double precision, intent(inout) :: v_0(sze_8,N_st), s_0(sze_8,N_st), u_0(sze_8,N_st)
integer :: k
do k=1,N_st
call dset_order(u_0(1,k),psi_bilinear_matrix_order,N_det)
enddo
call H_S2_u_0_nstates_bilinear_order(v_0,s_0,u_0,N_st,sze_8)
do k=1,N_st
call dset_order(v_0(1,k),psi_bilinear_matrix_order_reverse,N_det)
call dset_order(s_0(1,k),psi_bilinear_matrix_order_reverse,N_det)
call dset_order(u_0(1,k),psi_bilinear_matrix_order_reverse,N_det)
enddo
end
subroutine H_S2_u_0_nstates_bilinear_order(v_0,s_0,u_0,N_st,sze_8)
use bitmasks
implicit none
BEGIN_DOC
! Computes v_0 = H|u_0> and s_0 = S^2 |u_0>
END_DOC
integer, intent(in) :: N_st,sze_8
double precision, intent(out) :: v_0(sze_8,N_st), s_0(sze_8,N_st), u_0(sze_8,N_st)
Debug in davidson
2017-04-01 00:14:09 +02:00
PROVIDE ref_bitmask_energy
Bug corrected in selection tasks
2017-04-14 12:04:21 +02:00
double precision :: hij, sij
integer :: i,j,k,l
Debug in davidson
2017-04-01 00:14:09 +02:00
integer :: k_a, k_b, l_a, l_b, m_a, m_b
working on davidson
2017-04-14 11:09:55 +02:00
integer :: istate
Debug in davidson
2017-04-01 00:14:09 +02:00
integer :: krow, kcol, krow_b, kcol_b
integer :: lrow, lcol
integer :: mrow, mcol
integer(bit_kind) :: spindet(N_int)
integer(bit_kind) :: tmp_det(N_int,2)
integer(bit_kind) :: tmp_det2(N_int,2)
integer(bit_kind) :: tmp_det3(N_int,2)
integer(bit_kind), allocatable :: buffer(:,:)
integer :: n_singles, n_doubles
integer, allocatable :: singles(:), doubles(:)
OpenMP davidson
2017-04-14 16:40:12 +02:00
integer, allocatable :: singles_b(:,:)
Debug in davidson
2017-04-01 00:14:09 +02:00
integer, allocatable :: idx(:), idx0(:)
logical, allocatable :: is_single_a(:)
OpenMP davidson
2017-04-14 16:40:12 +02:00
integer :: maxab, n_singles_max, kcol_prev
New Davidson OK
2017-04-14 15:04:29 +02:00
double precision, allocatable :: u_t(:,:), v_t(:,:), s_t(:,:)
New Davidson OK
2017-04-14 15:41:35 +02:00
!DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: v_t, s_t, u_t
Debug in davidson
2017-04-01 00:14:09 +02:00
OpenMP davidson
2017-04-14 18:11:02 +02:00
maxab = max(N_det_alpha_unique, N_det_beta_unique)+1
allocate(idx0(maxab), u_t(N_st,N_det) )
Debug in davidson
2017-04-01 00:14:09 +02:00
Fast davidson
2017-04-14 11:18:13 +02:00
do i=1,maxab
idx0(i) = i
enddo
New Davidson OK
2017-04-14 15:41:35 +02:00
call dtranspose( &
u_0, &
size(u_0, 1), &
u_t, &
size(u_t, 1), &
N_det, N_st)
New Davidson OK
2017-04-14 15:04:29 +02:00
OpenMP davidson
2017-04-14 18:11:02 +02:00
v_0 = 0.d0
s_0 = 0.d0
Fast davidson
2017-04-14 11:18:13 +02:00
! Prepare the array of all alpha single excitations
! -------------------------------------------------
OpenMP davidson
2017-04-14 16:40:12 +02:00
!$OMP PARALLEL DEFAULT(NONE) &
!$OMP SHARED(psi_bilinear_matrix_rows, N_det, &
!$OMP psi_bilinear_matrix_columns, &
!$OMP psi_det_alpha_unique, psi_det_beta_unique, &
!$OMP n_det_alpha_unique, n_det_beta_unique, N_int, &
!$OMP psi_bilinear_matrix_transp_rows, &
!$OMP psi_bilinear_matrix_transp_columns, &
!$OMP psi_bilinear_matrix_transp_order, N_st, &
!$OMP psi_bilinear_matrix_order_transp_reverse, &
!$OMP singles_alpha, psi_bilinear_matrix_columns_loc, &
OpenMP davidson
2017-04-14 18:11:02 +02:00
!$OMP singles_alpha_size, sze_8, &
!$OMP idx0, u_t, maxab, v_0, s_0) &
OpenMP davidson
2017-04-14 16:40:12 +02:00
!$OMP PRIVATE(krow, kcol, tmp_det, spindet, k_a, k_b, i, &
!$OMP lcol, lrow, is_single_a,l_a, l_b, &
!$OMP buffer, singles, doubles, n_singles, n_doubles, &
OpenMP davidson
2017-04-14 18:11:02 +02:00
!$OMP tmp_det2, hij, sij, idx, l, kcol_prev, v_t, s_t)
OpenMP davidson
2017-04-14 16:40:12 +02:00
! Alpha/Beta double excitations
! =============================
allocate( buffer(N_int,maxab), &
OpenMP davidson
2017-04-14 18:11:02 +02:00
singles(singles_alpha_size), &
doubles(maxab), &
idx(maxab), &
v_t(N_st,N_det), s_t(N_st,N_det), &
OpenMP davidson
2017-04-14 16:40:12 +02:00
is_single_a(N_det_alpha_unique))
is_single_a = .False.
kcol_prev=-1
krow=1
OpenMP davidson
2017-04-14 18:11:02 +02:00
v_t = 0.d0
s_t = 0.d0
OpenMP davidson
2017-04-14 16:40:12 +02:00
!$OMP DO SCHEDULE(static,1)
do k_a=1,N_det
do k=1,singles_alpha(0,krow)
is_single_a( singles_alpha(k,krow) ) = .False.
enddo
krow = psi_bilinear_matrix_rows(k_a)
kcol = psi_bilinear_matrix_columns(k_a)
tmp_det(1:N_int,1) = psi_det_alpha_unique(1:N_int, krow)
tmp_det(1:N_int,2) = psi_det_beta_unique (1:N_int, kcol)
do k=1,singles_alpha(0,krow)
is_single_a( singles_alpha(k,krow) ) = .True.
enddo
if (kcol /= kcol_prev) then
call get_all_spin_singles( &
psi_det_beta_unique, idx0, tmp_det(1,2), N_int, N_det_beta_unique,&
singles, n_singles)
endif
kcol_prev = kcol
! Loop over singly excited beta columns
! -------------------------------------
do i=1,n_singles
lcol = singles(i)
if (lcol <= kcol) cycle
tmp_det2(1:N_int,2) = psi_det_beta_unique(1:N_int, lcol)
l_a = psi_bilinear_matrix_columns_loc(lcol)
do while (l_a <= k_a)
l_a += 1
enddo
n_doubles=1
do while ( l_a < psi_bilinear_matrix_columns_loc(lcol+1) )
lrow = psi_bilinear_matrix_rows(l_a)
if (.not.is_single_a(lrow)) then
continue
else
doubles(n_doubles) = lrow
idx(n_doubles) = l_a
OpenMP davidson
2017-04-14 18:11:02 +02:00
if (n_doubles == maxab) then
do k=1,n_doubles
lrow = doubles(k)
l_a = idx(k)
tmp_det2(1:N_int,1) = psi_det_alpha_unique(1:N_int, lrow)
call i_H_j_double_alpha_beta(tmp_det,tmp_det2,N_int,hij)
call get_s2(tmp_det,tmp_det2,N_int,sij)
do l=1,N_st
v_t(l,k_a) = v_t(l,k_a) + hij * u_t(l,l_a)
s_t(l,k_a) = s_t(l,k_a) + sij * u_t(l,l_a)
v_t(l,l_a) = v_t(l,l_a) + hij * u_t(l,k_a)
s_t(l,l_a) = s_t(l,l_a) + sij * u_t(l,k_a)
enddo
enddo
n_doubles=0
endif
OpenMP davidson
2017-04-14 16:40:12 +02:00
n_doubles = n_doubles+1
endif
l_a = l_a+1
enddo
n_doubles = n_doubles-1
OpenMP davidson
2017-04-14 18:11:02 +02:00
if (n_doubles > 0) then
do k=1,n_doubles
lrow = doubles(k)
l_a = idx(k)
tmp_det2(1:N_int,1) = psi_det_alpha_unique(1:N_int, lrow)
call i_H_j_double_alpha_beta(tmp_det,tmp_det2,N_int,hij)
call get_s2(tmp_det,tmp_det2,N_int,sij)
do l=1,N_st
v_t(l,k_a) = v_t(l,k_a) + hij * u_t(l,l_a)
s_t(l,k_a) = s_t(l,k_a) + sij * u_t(l,l_a)
v_t(l,l_a) = v_t(l,l_a) + hij * u_t(l,k_a)
s_t(l,l_a) = s_t(l,l_a) + sij * u_t(l,k_a)
enddo
enddo
endif
OpenMP davidson
2017-04-14 16:40:12 +02:00
enddo
! Diagonal contribution
! ---------------------
double precision, external :: diag_H_mat_elem, diag_S_mat_elem
hij = diag_H_mat_elem(tmp_det,N_int)
sij = diag_S_mat_elem(tmp_det,N_int)
do l=1,N_st
v_t(l,k_a) = v_t(l,k_a) + hij * u_t(l,k_a)
s_t(l,k_a) = s_t(l,k_a) + sij * u_t(l,k_a)
enddo
Fast davidson
2017-04-14 11:18:13 +02:00
enddo
OpenMP davidson
2017-04-14 16:40:12 +02:00
!$OMP END DO NOWAIT
Fast davidson
2017-04-14 11:18:13 +02:00
OpenMP davidson
2017-04-14 16:40:12 +02:00
!$OMP DO SCHEDULE(static,1)
working on davidson
2017-04-13 20:04:35 +02:00
do k_a=1,N_det
Debug in davidson
2017-04-01 00:14:09 +02:00
! Initial determinant is at k_a in alpha-major representation
! -----------------------------------------------------------------------
krow = psi_bilinear_matrix_rows(k_a)
kcol = psi_bilinear_matrix_columns(k_a)
tmp_det(1:N_int,1) = psi_det_alpha_unique(1:N_int, krow)
tmp_det(1:N_int,2) = psi_det_beta_unique (1:N_int, kcol)
! Initial determinant is at k_b in beta-major representation
! ----------------------------------------------------------------------
New Davidson OK
2017-04-14 15:04:29 +02:00
k_b = psi_bilinear_matrix_order_transp_reverse(k_a)
Same spin OK
2017-04-02 16:42:32 +02:00
Debug in davidson
2017-04-01 00:14:09 +02:00
! Get all single and double alpha excitations
! ===========================================
spindet(1:N_int) = tmp_det(1:N_int,1)
! Loop inside the beta column to gather all the connected alphas
i=1
New Davidson OK
2017-04-14 15:04:29 +02:00
l_a = k_a
Same spin OK
2017-04-02 16:42:32 +02:00
lcol = psi_bilinear_matrix_columns(l_a)
working on davidson
2017-04-14 11:09:55 +02:00
do while (lcol == kcol)
Debug in davidson
2017-04-01 00:14:09 +02:00
lrow = psi_bilinear_matrix_rows(l_a)
buffer(1:N_int,i) = psi_det_alpha_unique(1:N_int, lrow)
idx(i) = lrow
working on davidson
2017-04-14 11:09:55 +02:00
i = i +1
l_a = l_a+1
if (l_a > N_det) exit
Same spin OK
2017-04-02 16:42:32 +02:00
lcol = psi_bilinear_matrix_columns(l_a)
Debug in davidson
2017-04-01 00:14:09 +02:00
enddo
i = i-1
call get_all_spin_singles_and_doubles( &
buffer, idx, spindet, N_int, i, &
singles, doubles, n_singles, n_doubles )
! Compute Hij for all alpha singles
! ----------------------------------
Same spin OK
2017-04-02 16:42:32 +02:00
l_a = k_a
Debug in davidson
2017-04-01 00:14:09 +02:00
lrow = psi_bilinear_matrix_rows(l_a)
Same spin OK
2017-04-02 16:42:32 +02:00
tmp_det2(1:N_int,2) = psi_det_beta_unique (1:N_int, kcol)
Debug in davidson
2017-04-01 00:14:09 +02:00
do i=1,n_singles
do while ( lrow < singles(i) )
l_a = l_a+1
lrow = psi_bilinear_matrix_rows(l_a)
enddo
Same spin OK
2017-04-02 16:42:32 +02:00
tmp_det2(1:N_int,1) = psi_det_alpha_unique(1:N_int, lrow)
call i_H_j_mono_spin( tmp_det, tmp_det2, N_int, 1, hij)
Bug corrected in selection tasks
2017-04-14 12:04:21 +02:00
do l=1,N_st
New Davidson OK
2017-04-14 15:41:35 +02:00
v_t(l,l_a) = v_t(l,l_a) + hij * u_t(l,k_a)
v_t(l,k_a) = v_t(l,k_a) + hij * u_t(l,l_a)
Bug corrected in selection tasks
2017-04-14 12:04:21 +02:00
! single => sij = 0
enddo
Debug in davidson
2017-04-01 00:14:09 +02:00
enddo
! Compute Hij for all alpha doubles
! ----------------------------------
Same spin OK
2017-04-02 16:42:32 +02:00
l_a = k_a
Debug in davidson
2017-04-01 00:14:09 +02:00
lrow = psi_bilinear_matrix_rows(l_a)
do i=1,n_doubles
do while (lrow < doubles(i))
l_a = l_a+1
lrow = psi_bilinear_matrix_rows(l_a)
enddo
Same spin OK
2017-04-02 16:42:32 +02:00
call i_H_j_double_spin( tmp_det(1,1), psi_det_alpha_unique(1, doubles(i)), N_int, hij)
Bug corrected in selection tasks
2017-04-14 12:04:21 +02:00
do l=1,N_st
New Davidson OK
2017-04-14 15:41:35 +02:00
v_t(l,l_a) = v_t(l,l_a) + hij * u_t(l,k_a)
v_t(l,k_a) = v_t(l,k_a) + hij * u_t(l,l_a)
Bug corrected in selection tasks
2017-04-14 12:04:21 +02:00
! same spin => sij = 0
enddo
Debug in davidson
2017-04-01 00:14:09 +02:00
enddo
! Get all single and double beta excitations
! ===========================================
spindet(1:N_int) = tmp_det(1:N_int,2)
! Loop inside the alpha row to gather all the connected betas
i=1
New Davidson OK
2017-04-14 15:04:29 +02:00
l_b = k_b
Same spin OK
2017-04-02 16:42:32 +02:00
lrow = psi_bilinear_matrix_transp_rows(l_b)
working on davidson
2017-04-14 11:09:55 +02:00
do while (lrow == krow)
Debug in davidson
2017-04-01 00:14:09 +02:00
lcol = psi_bilinear_matrix_transp_columns(l_b)
buffer(1:N_int,i) = psi_det_beta_unique(1:N_int, lcol)
idx(i) = lcol
working on davidson
2017-04-14 11:09:55 +02:00
i = i +1
l_b = l_b+1
if (l_b > N_det) exit
Same spin OK
2017-04-02 16:42:32 +02:00
lrow = psi_bilinear_matrix_transp_rows(l_b)
Debug in davidson
2017-04-01 00:14:09 +02:00
enddo
i = i-1
call get_all_spin_singles_and_doubles( &
buffer, idx, spindet, N_int, i, &
singles, doubles, n_singles, n_doubles )
! Compute Hij for all beta singles
! ----------------------------------
l_b = k_b
lcol = psi_bilinear_matrix_transp_columns(l_b)
Same spin OK
2017-04-02 16:42:32 +02:00
tmp_det2(1:N_int,1) = psi_det_alpha_unique(1:N_int, krow)
Debug in davidson
2017-04-01 00:14:09 +02:00
do i=1,n_singles
do while ( lcol < singles(i) )
l_b = l_b+1
lcol = psi_bilinear_matrix_transp_columns(l_b)
enddo
Same spin OK
2017-04-02 16:42:32 +02:00
tmp_det2(1:N_int,2) = psi_det_beta_unique (1:N_int, lcol)
l_a = psi_bilinear_matrix_transp_order(l_b)
call i_H_j_mono_spin( tmp_det, tmp_det2, N_int, 2, hij)
Bug corrected in selection tasks
2017-04-14 12:04:21 +02:00
do l=1,N_st
New Davidson OK
2017-04-14 15:41:35 +02:00
v_t(l,l_a) = v_t(l,l_a) + hij * u_t(l,k_a)
v_t(l,k_a) = v_t(l,k_a) + hij * u_t(l,l_a)
Bug corrected in selection tasks
2017-04-14 12:04:21 +02:00
! single => sij = 0
enddo
Debug in davidson
2017-04-01 00:14:09 +02:00
enddo
! Compute Hij for all beta doubles
! ----------------------------------
l_b = k_b
lcol = psi_bilinear_matrix_transp_columns(l_b)
do i=1,n_doubles
do while (lcol < doubles(i))
l_b = l_b+1
lcol = psi_bilinear_matrix_transp_columns(l_b)
enddo
Same spin OK
2017-04-02 16:42:32 +02:00
l_a = psi_bilinear_matrix_transp_order(l_b)
call i_H_j_double_spin( tmp_det(1,2), psi_det_beta_unique(1, doubles(i)), N_int, hij)
Bug corrected in selection tasks
2017-04-14 12:04:21 +02:00
do l=1,N_st
New Davidson OK
2017-04-14 15:41:35 +02:00
v_t(l,l_a) = v_t(l,l_a) + hij * u_t(l,k_a)
v_t(l,k_a) = v_t(l,k_a) + hij * u_t(l,l_a)
Bug corrected in selection tasks
2017-04-14 12:04:21 +02:00
! same spin => sij = 0
enddo
Debug in davidson
2017-04-01 00:14:09 +02:00
enddo
end do
OpenMP davidson
2017-04-14 18:11:02 +02:00
!$OMP END DO NOWAIT
New Davidson OK
2017-04-14 15:04:29 +02:00
OpenMP davidson
2017-04-14 18:11:02 +02:00
!$OMP CRITICAL
do l=1,N_st
do i=1, N_det
v_0(i,l) = v_0(i,l) + v_t(l,i)
s_0(i,l) = s_0(i,l) + s_t(l,i)
enddo
enddo
!$OMP END CRITICAL
New Davidson OK
2017-04-14 15:41:35 +02:00
OpenMP davidson
2017-04-14 18:16:31 +02:00
!$OMP BARRIER
OpenMP davidson
2017-04-14 18:11:02 +02:00
!$OMP END PARALLEL
working on davidson
2017-04-14 11:09:55 +02:00
Debug in davidson
2017-04-01 00:14:09 +02:00
end
subroutine H_S2_u_0_nstates_test(v_0,s_0,u_0,H_jj,S2_jj,n,keys_tmp,Nint,N_st,sze_8)
use bitmasks
implicit none
integer, intent(in) :: N_st,n,Nint, sze_8
pedantic changes
2017-04-12 19:29:21 +02:00
integer(bit_kind), intent(in) :: keys_tmp(Nint,2,n)
Debug in davidson
2017-04-01 00:14:09 +02:00
double precision, intent(out) :: v_0(sze_8,N_st), s_0(sze_8,N_st)
double precision, intent(in) :: u_0(sze_8,N_st)
double precision, intent(in) :: H_jj(n), S2_jj(n)
PROVIDE ref_bitmask_energy
double precision, allocatable :: vt(:,:)
integer, allocatable :: idx(:)
integer :: i,j, jj
double precision :: hij
do i=1,n
v_0(i,:) = H_jj(i) * u_0(i,:)
enddo
allocate(idx(0:n), vt(N_st,n))
Vt = 0.d0
Same spin OK
2017-04-02 16:42:32 +02:00
do i=2,n
Debug in davidson
2017-04-01 00:14:09 +02:00
idx(0) = i
call filter_connected(keys_tmp,keys_tmp(1,1,i),Nint,i-1,idx)
do jj=1,idx(0)
j = idx(jj)
Same spin OK
2017-04-02 16:42:32 +02:00
double precision :: phase
integer :: degree
integer :: exc(0:2,2,2)
call get_excitation(keys_tmp(1,1,j),keys_tmp(1,1,i),exc,degree,phase,Nint)
working on davidson
2017-04-14 11:09:55 +02:00
! if ((degree == 2).and.(exc(0,1,1)==1)) cycle
working on davidson
2017-04-13 20:04:35 +02:00
! if ((degree > 1)) cycle
Same spin OK
2017-04-02 16:42:32 +02:00
! if (exc(0,1,2) /= 0) cycle
Debug in davidson
2017-04-01 00:14:09 +02:00
call i_H_j(keys_tmp(1,1,j),keys_tmp(1,1,i),Nint,hij)
vt (:,i) = vt (:,i) + hij*u_0(j,:)
vt (:,j) = vt (:,j) + hij*u_0(i,:)
enddo
enddo
do i=1,n
v_0(i,:) = v_0(i,:) + vt(:,i)
enddo
end
Reference in New Issue Copy Permalink