PTEROSOR/QuantumPackage
10
0
Fork 0
mirror of https://github.com/QuantumPackage/qp2.git synced 2025-03-31 14:01:36 +02:00
Code Releases Activity

Implemented single precision for cholesky mo

Browse Source
This commit is contained in:
Anthony Scemama 2025-01-24 19:07:16 +01:00
parent 081df88585
commit 0c6d513bbf
5 changed files with 198 additions and 56 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
src/ao_two_e_ints/cholesky.irp.f
View File

@ -466,10 +466,11 @@ END_PROVIDER
endif
! Reverse order of Cholesky vectors to increase precision in dot products
!$OMP PARALLEL DO PRIVATE(k,j)
do k=1,rank
do j=1,ao_num
cholesky_ao(1:ao_num,j,k) = L((j-1_8)*ao_num+1_8:1_8*j*ao_num,k)
cholesky_ao(1:ao_num,j,rank-k+1) = L((j-1_8)*ao_num+1_8:1_8*j*ao_num,rank-k+1)
enddo
enddo
!$OMP END PARALLEL DO

2
src/fci/pt2.irp.f
View File

@ -15,11 +15,11 @@ program pt2
! sampling.
!
END_DOC
PROVIDE all_mo_integrals
if (.not. is_zmq_slave) then
read_wf = .True.
threshold_generators = 1.d0
SOFT_TOUCH read_wf threshold_generators
PROVIDE all_mo_integrals
PROVIDE psi_energy
call run
else

59
src/mo_two_e_ints/cholesky.irp.f
View File

@ -7,7 +7,8 @@ BEGIN_PROVIDER [ logical, do_mo_cholesky ]
! do_mo_cholesky = .False.
END_PROVIDER
BEGIN_PROVIDER [ integer, cholesky_mo_num ]
BEGIN_PROVIDER [ integer, cholesky_mo_num ]
&BEGIN_PROVIDER [ integer, cholesky_mo_num_split, (1:5)]
implicit none
BEGIN_DOC
! Number of Cholesky vectors in MO basis
@ -21,6 +22,12 @@ BEGIN_PROVIDER [ integer, cholesky_mo_num ]
else
cholesky_mo_num = cholesky_ao_num
endif
cholesky_mo_num_split(1) = 0
cholesky_mo_num_split(2) = cholesky_mo_num/4
cholesky_mo_num_split(3) = 2*cholesky_mo_num_split(2)
cholesky_mo_num_split(4) = 3*cholesky_mo_num_split(2)
cholesky_mo_num_split(5) = cholesky_mo_num
cholesky_mo_num_split += 1
END_PROVIDER
BEGIN_PROVIDER [ double precision, cholesky_mo, (mo_num, mo_num, cholesky_mo_num) ]
@ -49,7 +56,7 @@ BEGIN_PROVIDER [ double precision, cholesky_mo_transp, (cholesky_mo_num, mo_num,
BEGIN_DOC
! Cholesky vectors in MO basis. Warning: it is transposed wrt cholesky_ao:
!
! - cholesky_ao is (ao_num^2 x cholesky_ao_num)
! - cholesky_ao is (ao_num^2 x cholesky_ao_num)
!
! - cholesky_mo_transp is (cholesky_mo_num x mo_num^2)
END_DOC
@ -132,3 +139,51 @@ BEGIN_PROVIDER [ double precision, cholesky_semi_mo_transp_simple, (cholesky_mo_
END_PROVIDER
BEGIN_PROVIDER [ real, cholesky_mo_sp, (mo_num, mo_num, cholesky_mo_num) ]
implicit none
BEGIN_DOC
! Cholesky vectors in MO basis in stored in single precision
END_DOC
integer :: k, i, j
call set_multiple_levels_omp(.False.)
!$OMP PARALLEL DO PRIVATE(k)
do k=1,cholesky_mo_num
do j=1,mo_num
do i=1,mo_num
cholesky_mo_sp(i,j,k) = cholesky_mo_transp_sp(k,i,j)
enddo
enddo
enddo
!$OMP END PARALLEL DO
END_PROVIDER
BEGIN_PROVIDER [ real, cholesky_mo_transp_sp, (cholesky_mo_num, mo_num, mo_num) ]
implicit none
BEGIN_DOC
! Cholesky vectors in MO basis in s. Warning: it is transposed wrt cholesky_ao:
!
! - cholesky_ao is (ao_num^2 x cholesky_ao_num)
!
! - cholesky_mo_transp is (cholesky_mo_num x mo_num^2)
END_DOC
integer :: i,j,k
!$OMP PARALLEL DO PRIVATE(k)
do k=1,cholesky_mo_num
do j=1,mo_num
do i=1,mo_num
cholesky_mo_transp_sp(k,i,j) = cholesky_mo_transp(k,i,j)
enddo
enddo
enddo
!$OMP END PARALLEL DO
END_PROVIDER

188
src/mo_two_e_ints/map_integrals.irp.f
View File

@ -9,6 +9,14 @@ BEGIN_PROVIDER [ logical, all_mo_integrals ]
PROVIDE mo_two_e_integrals_in_map mo_integrals_cache mo_two_e_integrals_jj_exchange mo_two_e_integrals_jj_anti mo_two_e_integrals_jj big_array_exchange_integrals big_array_coulomb_integrals mo_one_e_integrals
END_PROVIDER
BEGIN_PROVIDER [ logical, mo_cholesky_double ]
implicit none
BEGIN_DOC
! If true, use double precision to compute integrals from cholesky vectors
END_DOC
mo_cholesky_double = .True.
END_PROVIDER
!! MO Map
!! ======
@ -147,7 +155,7 @@ double precision function get_two_e_integral(i,j,k,l,map)
type(map_type), intent(inout) :: map
real(integral_kind) :: tmp
PROVIDE mo_two_e_integrals_in_map mo_integrals_cache do_mo_cholesky
PROVIDE mo_two_e_integrals_in_map mo_integrals_cache do_mo_cholesky mo_cholesky_double cholesky_mo_transp_sp cholesky_mo_transp
if (use_banned_excitation) then
if (banned_excitation(i,k)) then
@ -178,12 +186,19 @@ double precision function get_two_e_integral(i,j,k,l,map)
if (do_mo_cholesky) then
double precision, external :: ddot
get_two_e_integral = ddot(cholesky_mo_num, cholesky_mo_transp(1,i,k), 1, cholesky_mo_transp(1,j,l), 1)
! get_two_e_integral = 0.d0
! do kk=1,cholesky_mo_num
! get_two_e_integral = get_two_e_integral + cholesky_mo_transp(kk,i,k)*cholesky_mo_transp(kk,i,l)
! enddo
real, external :: sdot
integer :: isplit
if (mo_cholesky_double) then
get_two_e_integral = ddot(cholesky_mo_num, cholesky_mo_transp(1,i,k), 1, cholesky_mo_transp(1,j,l), 1)
else
get_two_e_integral = 0.d0
do isplit=1,4
get_two_e_integral = get_two_e_integral + &
sdot(cholesky_mo_num_split(isplit+1) - cholesky_mo_num_split(isplit), &
cholesky_mo_transp_sp(cholesky_mo_num_split(isplit),i,k), 1, &
cholesky_mo_transp_sp(cholesky_mo_num_split(isplit),j,l), 1)
enddo
endif
else
@ -214,7 +229,8 @@ subroutine get_mo_two_e_integrals(j,k,l,sze,out_val,map)
real(integral_kind) :: tmp
integer(key_kind) :: i1, idx
integer(key_kind) :: p,q,r,s,i2
PROVIDE mo_two_e_integrals_in_map mo_integrals_cache
real, allocatable :: out_val_sp(:)
PROVIDE mo_two_e_integrals_in_map mo_integrals_cache cholesky_mo_transp cholesky_mo_transp_sp
if (banned_excitation(j,l)) then
out_val(1:sze) = 0.d0
@ -225,6 +241,10 @@ subroutine get_mo_two_e_integrals(j,k,l,sze,out_val,map)
ii = ior(ii, k-mo_integrals_cache_min)
ii = ior(ii, j-mo_integrals_cache_min)
if (do_mo_cholesky.and. .not.mo_cholesky_double) then
allocate(out_val_sp(sze))
endif
if (iand(ii, -mo_integrals_cache_size) == 0) then
! Some integrals are in the cache
@ -232,11 +252,24 @@ subroutine get_mo_two_e_integrals(j,k,l,sze,out_val,map)
if (do_mo_cholesky) then
!TODO: here
call dgemv('T', cholesky_mo_num, mo_integrals_cache_min-1, 1.d0, &
cholesky_mo_transp(1,1,k), cholesky_mo_num, &
cholesky_mo_transp(1,j,l), 1, 0.d0, &
out_val, 1)
!TODO: bottleneck here
if (mo_cholesky_double) then
call dgemv('T', cholesky_mo_num, mo_integrals_cache_min-1, 1.d0, &
cholesky_mo_transp(1,1,k), cholesky_mo_num, &
cholesky_mo_transp(1,j,l), 1, 0.d0, &
out_val, 1)
else
integer :: isplit
out_val = 0.d0
do isplit=1,4
call sgemv('T', cholesky_mo_num_split(isplit+1) - cholesky_mo_num_split(isplit), &
mo_integrals_cache_min-1, 1., &
cholesky_mo_transp_sp(cholesky_mo_num_split(isplit),1,k), cholesky_mo_num, &
cholesky_mo_transp_sp(cholesky_mo_num_split(isplit),j,l), 1, 0., &
out_val_sp, 1)
out_val(1:mo_integrals_cache_min-1) += out_val_sp(1:mo_integrals_cache_min-1)
enddo
endif
else
@ -270,11 +303,23 @@ subroutine get_mo_two_e_integrals(j,k,l,sze,out_val,map)
if (do_mo_cholesky) then
!TODO: here
call dgemv('T', cholesky_mo_num, mo_num-mo_integrals_cache_max, 1.d0, &
cholesky_mo_transp(1,mo_integrals_cache_max+1,k), cholesky_mo_num, &
cholesky_mo_transp(1,j,l), 1, 0.d0, &
out_val(mo_integrals_cache_max+1), 1)
!TODO: bottleneck here
if (mo_cholesky_double) then
call dgemv('T', cholesky_mo_num, mo_num-mo_integrals_cache_max, 1.d0, &
cholesky_mo_transp(1,mo_integrals_cache_max+1,k), cholesky_mo_num, &
cholesky_mo_transp(1,j,l), 1, 0.d0, &
out_val(mo_integrals_cache_max+1), 1)
else
out_val = 0.d0
do isplit=1,4
call sgemv('T', cholesky_mo_num_split(isplit+1) - cholesky_mo_num_split(isplit), &
mo_num-mo_integrals_cache_max, 1., &
cholesky_mo_transp_sp(cholesky_mo_num_split(isplit),mo_integrals_cache_max+1,k), cholesky_mo_num, &
cholesky_mo_transp_sp(cholesky_mo_num_split(isplit),j,l), 1, 0., &
out_val_sp(mo_integrals_cache_max+1), 1)
out_val(mo_integrals_cache_max+1:sze) += out_val_sp(mo_integrals_cache_max+1:sze)
enddo
endif
else
@ -306,11 +351,23 @@ subroutine get_mo_two_e_integrals(j,k,l,sze,out_val,map)
if (do_mo_cholesky) then
!TODO: here
call dgemv('T', cholesky_mo_num, mo_num, 1.d0, &
cholesky_mo_transp(1,1,k), cholesky_mo_num, &
cholesky_mo_transp(1,j,l), 1, 0.d0, &
out_val, 1)
!TODO: bottleneck here
if (mo_cholesky_double) then
call dgemv('T', cholesky_mo_num, sze, 1.d0, &
cholesky_mo_transp(1,1,k), cholesky_mo_num, &
cholesky_mo_transp(1,j,l), 1, 0.d0, &
out_val, 1)
else
out_val = 0.d0
do isplit=1,4
call sgemv('T', cholesky_mo_num_split(isplit+1) - cholesky_mo_num_split(isplit), &
sze, 1., &
cholesky_mo_transp_sp(cholesky_mo_num_split(isplit),1,k), cholesky_mo_num, &
cholesky_mo_transp_sp(cholesky_mo_num_split(isplit),j,l), 1, 0., &
out_val_sp, 1)
out_val(1:sze) += out_val_sp(1:sze)
enddo
endif
else
@ -513,7 +570,7 @@ subroutine get_mo_two_e_integrals_exch_ii(k,l,sze,out_val,map)
type(map_type), intent(inout) :: map
integer :: i
double precision, external :: get_two_e_integral
PROVIDE mo_two_e_integrals_in_map
PROVIDE mo_two_e_integrals_in_map mo_cholesky_double cholesky_mo_transp cholesky_mo_transp_sp
if ( (mo_integrals_cache_min>1).or.(mo_integrals_cache_max<mo_num) ) then
@ -523,40 +580,69 @@ subroutine get_mo_two_e_integrals_exch_ii(k,l,sze,out_val,map)
(l>=mo_integrals_cache_min).and.(l<=mo_integrals_cache_max) ) then
double precision, external :: ddot
real, external :: sdot
integer :: kk
do i=1,mo_integrals_cache_min-1
out_val(i) = ddot(cholesky_mo_num, cholesky_mo_transp(1,i,k), 1, &
cholesky_mo_transp(1,i,l), 1)
! out_val(i) = 0.d0
! do kk=1,cholesky_mo_num
! out_val(i) = out_val(i) + cholesky_mo_transp(kk,i,k)*cholesky_mo_transp(kk,i,l)
! enddo
enddo
if (mo_cholesky_double) then
do i=mo_integrals_cache_min,mo_integrals_cache_max
out_val(i) = get_two_e_integral_cache(i,i,k,l)
enddo
do i=1,mo_integrals_cache_min-1
out_val(i) = ddot(cholesky_mo_num, cholesky_mo_transp(1,i,k), 1, &
cholesky_mo_transp(1,i,l), 1)
enddo
do i=mo_integrals_cache_max, sze
out_val(i) = ddot(cholesky_mo_num, cholesky_mo_transp(1,i,k), 1, &
cholesky_mo_transp(1,i,l), 1)
! out_val(i) = 0.d0
! do kk=1,cholesky_mo_num
! out_val(i) = out_val(i) + cholesky_mo_transp(kk,i,k)*cholesky_mo_transp(kk,i,l)
! enddo
enddo
do i=mo_integrals_cache_min,mo_integrals_cache_max
out_val(i) = get_two_e_integral_cache(i,i,k,l)
enddo
do i=mo_integrals_cache_max, sze
out_val(i) = ddot(cholesky_mo_num, cholesky_mo_transp(1,i,k), 1, &
cholesky_mo_transp(1,i,l), 1)
enddo
else
integer :: isplit
do i=1,mo_integrals_cache_min-1
out_val(i) = 0.d0
do isplit=1,4
out_val(i) += sdot(cholesky_mo_num_split(isplit+1) - cholesky_mo_num_split(isplit), &
cholesky_mo_transp_sp(cholesky_mo_num_split(isplit),i,k), 1, &
cholesky_mo_transp_sp(cholesky_mo_num_split(isplit),i,l), 1)
enddo
enddo
do i=mo_integrals_cache_min,mo_integrals_cache_max
out_val(i) = get_two_e_integral_cache(i,i,k,l)
enddo
do i=mo_integrals_cache_max, sze
out_val(i) = 0.d0
do isplit=1,4
out_val(i) += sdot(cholesky_mo_num_split(isplit+1) - cholesky_mo_num_split(isplit), &
cholesky_mo_transp_sp(cholesky_mo_num_split(isplit),i,k), 1, &
cholesky_mo_transp_sp(cholesky_mo_num_split(isplit),i,l), 1)
enddo
enddo
endif
else
do i=1,sze
out_val(i) = ddot(cholesky_mo_num, cholesky_mo_transp(1,i,k), 1, &
cholesky_mo_transp(1,i,l), 1)
! out_val(i) = 0.d0
! do kk=1,cholesky_mo_num
! out_val(i) = out_val(i) + cholesky_mo_transp(kk,i,k)*cholesky_mo_transp(kk,i,l)
! enddo
enddo
if (mo_cholesky_double) then
do i=1,sze
out_val(i) = ddot(cholesky_mo_num, cholesky_mo_transp(1,i,k), 1, &
cholesky_mo_transp(1,i,l), 1)
enddo
else
do i=1,sze
out_val(i) = 0.d0
do isplit=1,4
out_val(i) += sdot(cholesky_mo_num_split(isplit+1) - cholesky_mo_num_split(isplit), &
cholesky_mo_transp_sp(cholesky_mo_num_split(isplit),i,k), 1, &
cholesky_mo_transp_sp(cholesky_mo_num_split(isplit),i,l), 1)
enddo
enddo
endif
endif

2
src/mp2/mp2.irp.f
View File

@ -7,7 +7,7 @@ subroutine run
double precision, allocatable :: pt2(:), norm_pert(:)
double precision :: H_pert_diag, E_old
integer :: N_st, iter
PROVIDE Fock_matrix_diag_mo H_apply_buffer_allocated
PROVIDE all_mo_integrals Fock_matrix_diag_mo H_apply_buffer_allocated
N_st = N_states
allocate (pt2(N_st), norm_pert(N_st))
E_old = HF_energy