10
0
mirror of https://github.com/LCPQ/quantum_package synced 2024-11-12 17:13:54 +01:00

Accelerated AO integrals

This commit is contained in:
Anthony Scemama 2014-09-26 19:12:09 +02:00
parent 9485610d5d
commit 609cc1b7a4

View File

@ -121,8 +121,8 @@ double precision function ao_bielec_integral_schwartz_accel(i,j,k,l)
double precision :: P_new(0:max_dim,3),P_center(3),fact_p,pp
double precision :: Q_new(0:max_dim,3),Q_center(3),fact_q,qq
integer :: iorder_p(3), iorder_q(3)
double precision, allocatable :: schwartz_kl(:,:)
double precision :: schwartz_ij
real, allocatable :: schwartz_kl(:,:)
real :: schwartz_ij
PROVIDE all_utils
@ -151,10 +151,10 @@ double precision function ao_bielec_integral_schwartz_accel(i,j,k,l)
L_center(p) = nucl_coord(num_l,p)
enddo
schwartz_kl(0,0) = 0.d0
schwartz_kl(0,0) = 0.
do r = 1, ao_prim_num(k)
coef1 = ao_coef_transp(r,k)*ao_coef_transp(r,k)
schwartz_kl(0,r) = 0.d0
schwartz_kl(0,r) = 0.
do s = 1, ao_prim_num(l)
coef2 = coef1 * ao_coef_transp(s,l) * ao_coef_transp(s,l)
call give_explicit_poly_and_gaussian(Q_new,Q_center,qq,fact_q,iorder_q,&
@ -205,10 +205,9 @@ double precision function ao_bielec_integral_schwartz_accel(i,j,k,l)
ao_expo_transp(r,k),ao_expo_transp(s,l), &
K_power,L_power,K_center,L_center,dim1)
q_inv = 1.d0/qq
integral = general_primitive_integral(dim1, &
integral += general_primitive_integral(dim1, &
P_new,P_center,fact_p,pp,p_inv,iorder_p, &
Q_new,Q_center,fact_q,qq,q_inv,iorder_q)
ao_bielec_integral_schwartz_accel += coef4 * integral
Q_new,Q_center,fact_q,qq,q_inv,iorder_q) * coef4
enddo ! s
enddo ! r
enddo ! q
@ -224,10 +223,10 @@ double precision function ao_bielec_integral_schwartz_accel(i,j,k,l)
enddo
double precision :: ERI
schwartz_kl(0,0) = 0.d0
schwartz_kl(0,0) = 0.
do r = 1, ao_prim_num(k)
coef1 = ao_coef_transp(r,k)*ao_coef_transp(r,k)
schwartz_kl(0,r) = 0.d0
schwartz_kl(0,r) = 0.
do s = 1, ao_prim_num(l)
coef2 = coef1*ao_coef_transp(s,l)*ao_coef_transp(s,l)
schwartz_kl(s,r) = ERI( &
@ -263,7 +262,7 @@ double precision function ao_bielec_integral_schwartz_accel(i,j,k,l)
cycle
endif
coef4 = coef3*ao_coef_transp(s,l)
integral = ERI( &
integral += coef4*ERI( &
ao_expo_transp(p,i),ao_expo_transp(q,j),ao_expo_transp(r,k),ao_expo_transp(s,l),&
I_power(1),J_power(1),K_power(1),L_power(1), &
I_power(2),J_power(2),K_power(2),L_power(2), &
@ -342,9 +341,8 @@ BEGIN_PROVIDER [ logical, ao_bielec_integrals_in_map ]
integer :: i,j,k,l
double precision :: ao_bielec_integral,cpu_1,cpu_2, wall_1, wall_2
double precision :: integral, wall_0
double precision :: thresh
thresh = ao_integrals_threshold
double precision :: wall_0
double precision :: thresh, thresh2
! For integrals file
integer*8,allocatable :: buffer_i(:)
@ -354,12 +352,37 @@ BEGIN_PROVIDER [ logical, ao_bielec_integrals_in_map ]
integer :: n_integrals, n_centers, thread_num
integer :: jl_pairs(2,ao_num*(ao_num+1)/2), kk, m, j1, i1, lmax
integer :: p,q,r,s
double precision :: I_center(3),J_center(3),K_center(3),L_center(3)
integer :: num_i,num_j,num_k,num_l,dim1,I_power(3),J_power(3),K_power(3),L_power(3)
double precision :: integral
include 'include/constants.F'
double precision :: P_new(0:max_dim,3),P_center(3),fact_p,pp
double precision :: Q_new(0:max_dim,3),Q_center(3),fact_q,qq
integer :: iorder_p(3), iorder_q(3)
real, allocatable :: schwartz_kl(:,:)
real :: schwartz_ij
double precision :: coef1
double precision :: coef2
double precision :: p_inv,q_inv
double precision :: coef3
double precision :: coef4
double precision :: general_primitive_integral
double precision :: ERI
thresh = ao_integrals_threshold
PROVIDE gauleg_t2 ao_integrals_map all_utils
! PROVIDE ao_bielec_integral_schwartz
PROVIDE ao_bielec_integral_schwartz
integral = ao_bielec_integral(1,1,1,1)
dim1 = n_pt_max_integrals
real :: map_mb
thresh2 = thresh*thresh
if (read_ao_integrals) then
integer :: load_ao_integrals
if (load_ao_integrals(trim(ezfio_filename)//'/work/ao_integrals.bin') == 0) then
@ -371,8 +394,8 @@ BEGIN_PROVIDER [ logical, ao_bielec_integrals_in_map ]
kk=1
do l = 1, ao_num ! r2
do j = 1, l ! r2
jl_pairs(1,kk) = j
do k = 1, l ! r2
jl_pairs(1,kk) = k
jl_pairs(2,kk) = l
kk += 1
enddo
@ -387,12 +410,17 @@ BEGIN_PROVIDER [ logical, ao_bielec_integrals_in_map ]
call cpu_time(cpu_1)
!$OMP PARALLEL PRIVATE(i,j,k,l,kk, &
!$OMP integral,buffer_i,buffer_value,n_integrals, &
!$OMP cpu_2,wall_2,i1,j1,thread_num) &
!$OMP cpu_2,wall_2,i1,j1,thread_num, num_i, num_j, num_k, num_l,&
!$OMP schwartz_kl,schwartz_ij,I_power,J_power,K_power,L_power, &
!$OMP I_center,J_center,K_center,L_center,coef1,coef2,coef3,coef4,&
!$OMP pp,qq,p_inv,q_inv,P_center,Q_center,P_new,Q_new,fact_p, &
!$OMP fact_q,iorder_p,iorder_q) &
!$OMP DEFAULT(NONE) &
!$OMP SHARED (ao_num, jl_pairs, ao_integrals_map,thresh, &
!$OMP cpu_1,wall_1,lock, lmax,n_centers,ao_nucl, &
!$OMP ao_overlap_abs,ao_overlap,output_BiInts,abort_here, &
!$OMP wall_0)
!$OMP SHARED (ao_num, jl_pairs, ao_integrals_map,thresh,thresh2,&
!$OMP cpu_1,wall_1,lock, lmax,n_centers,ao_nucl,nucl_coord, &
!$OMP ao_bielec_integral_schwartz,output_BiInts,abort_here, &
!$OMP wall_0,ao_prim_num,ao_expo_transp,ao_coef_transp,dim1, &
!$OMP ao_power)
allocate(buffer_i(size_buffer))
allocate(buffer_value(size_buffer))
@ -409,33 +437,198 @@ IRP_ENDIF
if (abort_here) then
cycle
endif
j = jl_pairs(1,kk)
k = jl_pairs(1,kk)
l = jl_pairs(2,kk)
j1 = j+ishft(l*l-l,-1)
if (ao_overlap_abs(j,l) < thresh) then
cycle
j1 = k+ishft(l*l-l,-1)
num_k = ao_nucl(k)
num_l = ao_nucl(l)
allocate(schwartz_kl(0:ao_prim_num(l),0:ao_prim_num(k)))
do p = 1, 3
K_power(p) = ao_power(k,p)
L_power(p) = ao_power(l,p)
K_center(p) = nucl_coord(num_k,p)
L_center(p) = nucl_coord(num_l,p)
enddo
! ====================================
! Schwartz screening for kl primitives
! ------------------------------------
schwartz_kl(0,0) = 0.
if (num_k /= num_l) then
do r = 1, ao_prim_num(k)
coef1 = ao_coef_transp(r,k)*ao_coef_transp(r,k)
schwartz_kl(0,r) = 0.
do s = 1, ao_prim_num(l)
coef2 = coef1 * ao_coef_transp(s,l) * ao_coef_transp(s,l)
call give_explicit_poly_and_gaussian(Q_new,Q_center,qq,fact_q,iorder_q,&
ao_expo_transp(r,k),ao_expo_transp(s,l), &
K_power,L_power,K_center,L_center,dim1)
q_inv = 1.d0/qq
schwartz_kl(s,r) = general_primitive_integral(dim1, &
Q_new,Q_center,fact_q,qq,q_inv,iorder_q, &
Q_new,Q_center,fact_q,qq,q_inv,iorder_q) &
* coef2
schwartz_kl(0,r) = max(schwartz_kl(0,r),schwartz_kl(s,r))
enddo
schwartz_kl(0,0) = max(schwartz_kl(0,r),schwartz_kl(0,0))
enddo
else
do r = 1, ao_prim_num(k)
coef1 = ao_coef_transp(r,k)*ao_coef_transp(r,k)
schwartz_kl(0,r) = 0.
do s = 1, ao_prim_num(l)
coef2 = coef1*ao_coef_transp(s,l)*ao_coef_transp(s,l)
schwartz_kl(s,r) = ERI( &
ao_expo_transp(r,k),ao_expo_transp(s,l),ao_expo_transp(r,k),ao_expo_transp(s,l),&
K_power(1),L_power(1),K_power(1),L_power(1), &
K_power(2),L_power(2),K_power(2),L_power(2), &
K_power(3),L_power(3),K_power(3),L_power(3)) * &
coef2
schwartz_kl(0,r) = max(schwartz_kl(0,r),schwartz_kl(s,r))
enddo
schwartz_kl(0,0) = max(schwartz_kl(0,r),schwartz_kl(0,0))
enddo
endif
do k = 1, ao_num ! r1
i1 = ishft(k*k-k,-1)
! -------------------------------------------
! End of Schwartz screening for kl primitives
! ===========================================
do j = 1, ao_num ! r1
i1 = ishft(j*j-j,-1)
if (i1 > j1) then
exit
endif
do i = 1, k
num_j = ao_nucl(j)
do p = 1, 3
J_power(p) = ao_power(j,p)
J_center(p) = nucl_coord(num_j,p)
enddo
do i = 1, j
i1 += 1
if (i1 > j1) then
exit
endif
if (ao_overlap_abs(i,k)*ao_overlap_abs(j,l) < thresh) then
cycle
num_i = ao_nucl(i)
do p = 1, 3
I_power(p) = ao_power(i,p)
I_center(p) = nucl_coord(num_i,p)
enddo
integral = 0.d0
if (num_i /= num_j .or. num_k /= num_l .or. num_j /= num_k)then
! =====================
! Multi-center integral
! ---------------------
do p = 1, ao_prim_num(i)
coef1 = ao_coef_transp(p,i)
do q = 1, ao_prim_num(j)
coef2 = coef1*ao_coef_transp(q,j)
call give_explicit_poly_and_gaussian(P_new,P_center,pp,fact_p,iorder_p,&
ao_expo_transp(p,i),ao_expo_transp(q,j), &
I_power,J_power,I_center,J_center,dim1)
p_inv = 1.d0/pp
schwartz_ij = general_primitive_integral(dim1, &
P_new,P_center,fact_p,pp,p_inv,iorder_p, &
P_new,P_center,fact_p,pp,p_inv,iorder_p) * &
coef2*coef2
if (schwartz_kl(0,0)*schwartz_ij < thresh2) then
cycle
endif
do r = 1, ao_prim_num(k)
if (schwartz_kl(0,r)*schwartz_ij < thresh2) then
cycle
endif
coef3 = coef2*ao_coef_transp(r,k)
do s = 1, ao_prim_num(l)
if (schwartz_kl(s,r)*schwartz_ij < thresh2) then
cycle
endif
coef4 = coef3*ao_coef_transp(s,l)
call give_explicit_poly_and_gaussian(Q_new,Q_center,qq,fact_q,iorder_q,&
ao_expo_transp(r,k),ao_expo_transp(s,l), &
K_power,L_power,K_center,L_center,dim1)
q_inv = 1.d0/qq
integral += coef4 * general_primitive_integral(dim1, &
P_new,P_center,fact_p,pp,p_inv,iorder_p, &
Q_new,Q_center,fact_q,qq,q_inv,iorder_q)
enddo ! s
enddo ! r
enddo ! q
enddo ! p
! -------------------------
! End Multi-center integral
! =========================
else
! ===================
! One-center integral
! -------------------
do p = 1, ao_prim_num(i)
coef1 = ao_coef_transp(p,i)
do q = 1, ao_prim_num(j)
coef2 = coef1*ao_coef_transp(q,j)
schwartz_ij = ERI( &
ao_expo_transp(p,i),ao_expo_transp(q,j),ao_expo_transp(p,i),ao_expo_transp(q,j),&
I_power(1),J_power(1),I_power(1),J_power(1), &
I_power(2),J_power(2),I_power(2),J_power(2), &
I_power(3),J_power(3),I_power(3),J_power(3))*coef2*coef2
if (schwartz_kl(0,0)*schwartz_ij < thresh2) then
cycle
endif
do r = 1, ao_prim_num(k)
if (schwartz_kl(0,r)*schwartz_ij < thresh2) then
cycle
endif
coef3 = coef2*ao_coef_transp(r,k)
do s = 1, ao_prim_num(l)
if (schwartz_kl(s,r)*schwartz_ij < thresh2) then
cycle
endif
coef4 = coef3*ao_coef_transp(s,l)
integral += coef4*ERI( &
ao_expo_transp(p,i),ao_expo_transp(q,j),ao_expo_transp(r,k),ao_expo_transp(s,l),&
I_power(1),J_power(1),K_power(1),L_power(1), &
I_power(2),J_power(2),K_power(2),L_power(2), &
I_power(3),J_power(3),K_power(3),L_power(3))
enddo ! s
enddo ! r
enddo ! q
enddo ! p
! -----------------------
! End One-center integral
! =======================
endif
!DIR$ FORCEINLINE
integral = ao_bielec_integral(i,k,j,l)
if (abs(integral) < thresh) then
cycle
endif
! ======================
! Append integral to map
! ----------------------
n_integrals += 1
!DIR$ FORCEINLINE
call bielec_integrals_index(i,j,k,l,buffer_i(n_integrals))
call bielec_integrals_index(i,k,j,l,buffer_i(n_integrals))
buffer_value(n_integrals) = integral
if (n_integrals > 1024 ) then
if (omp_test_lock(lock)) then
@ -448,6 +641,11 @@ IRP_ENDIF
call insert_into_ao_integrals_map(n_integrals,buffer_i,buffer_value)
n_integrals = 0
endif
! --------------------------
! End Append integral to map
! ==========================
enddo
enddo
call wall_time(wall_2)
@ -459,6 +657,7 @@ IRP_ENDIF
wall_2-wall_1, 's', map_mb(ao_integrals_map) ,'MB'
endif
endif
deallocate (schwartz_kl)
enddo
!$OMP END DO NOWAIT
call insert_into_ao_integrals_map(n_integrals,buffer_i,buffer_value)
@ -495,7 +694,7 @@ END_PROVIDER
BEGIN_PROVIDER [ double precision, ao_bielec_integral_schwartz,(ao_num,ao_num) ]
BEGIN_PROVIDER [ real, ao_bielec_integral_schwartz,(ao_num,ao_num) ]
implicit none
BEGIN_DOC
! Needed to compute Schwartz inequalities
@ -512,7 +711,7 @@ BEGIN_PROVIDER [ double precision, ao_bielec_integral_schwartz,(ao_num,ao_num)
!$OMP SCHEDULE(dynamic)
do i=1,ao_num
do k=1,i
ao_bielec_integral_schwartz(i,k) = dsqrt(ao_bielec_integral(i,k,i,k))
ao_bielec_integral_schwartz(i,k) = sqrt(ao_bielec_integral(i,k,i,k))
ao_bielec_integral_schwartz(k,i) = ao_bielec_integral_schwartz(i,k)
enddo
enddo
@ -807,8 +1006,7 @@ recursive subroutine I_x1_new(a,c,B_10,B_01,B_00,res,n_pt)
call I_x1_new(a-1,c-1,B_10,B_01,B_00,res2,n_pt)
!DIR$ VECTOR ALIGNED
do i=1,n_pt
res(i) = (a-1) * B_10(i) * res(i) &
+ c * B_00(i) * res2(i)
res(i) = (a-1) * B_10(i) * res(i) + c * B_00(i) * res2(i)
enddo
endif
end