LCPQ/qmcchem
10
1
Fork 1
mirror of https://gitlab.com/scemama/qmcchem.git synced 2024-11-07 06:33:38 +01:00
Code Releases Activity

Removed padding in sparse_full_mv

Browse Source
This commit is contained in:
Anthony Scemama 2016-06-03 14:14:27 +02:00
parent 033025c0ea
commit 052b2db389
3 changed files with 40 additions and 38 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
src/AO/ao_axis.irp.f
View File

@ -48,11 +48,11 @@ subroutine pow_l(r,a,x1,x2,x3)
end function end function
BEGIN_PROVIDER [ real, ao_axis_block, ((-2*simd_sp+1):ao_block_num_8) ] BEGIN_PROVIDER [ real, ao_axis_block, (ao_block_num_8) ]
&BEGIN_PROVIDER [ real, ao_axis_grad_block_x, ((-2*simd_sp+1):ao_block_num_8) ] &BEGIN_PROVIDER [ real, ao_axis_grad_block_x, (ao_block_num_8) ]
&BEGIN_PROVIDER [ real, ao_axis_grad_block_y, ((-2*simd_sp+1):ao_block_num_8) ] &BEGIN_PROVIDER [ real, ao_axis_grad_block_y, (ao_block_num_8) ]
&BEGIN_PROVIDER [ real, ao_axis_grad_block_z, ((-2*simd_sp+1):ao_block_num_8) ] &BEGIN_PROVIDER [ real, ao_axis_grad_block_z, (ao_block_num_8) ]
&BEGIN_PROVIDER [ real, ao_axis_lapl_block, ((-2*simd_sp+1):ao_block_num_8) ] &BEGIN_PROVIDER [ real, ao_axis_lapl_block, (ao_block_num_8) ]
implicit none implicit none
include '../types.F' include '../types.F'
@ -111,13 +111,13 @@ end function
ao_axis_block(idx) = p023 * p10 ao_axis_block(idx) = p023 * p10
p023 = real_of_int(pow1) * p023 p023 = real_of_int(pow1) * p023
ao_axis_grad_block_x(idx) = p023 * p11
ao_axis_grad_block_y(idx) = p013 * p21
ao_axis_grad_block_z(idx) = p012 * p31
ao_axis_lapl_block(idx) = real_of_int(pow1-1) * p023 * p12 & ao_axis_lapl_block(idx) = real_of_int(pow1-1) * p023 * p12 &
+ real_of_int(pow2-1) * p013 * p22 & + real_of_int(pow2-1) * p013 * p22 &
+ real_of_int(pow3-1) * p012 * p32 + real_of_int(pow3-1) * p012 * p32
ao_axis_grad_block_x(idx) = p023 * p11
ao_axis_grad_block_y(idx) = p013 * p21
ao_axis_grad_block_z(idx) = p012 * p31
enddo enddo

4
src/electrons.irp.f
View File

@ -148,8 +148,8 @@ END_PROVIDER
BEGIN_PROVIDER [ real, elec_dist, (elec_num_8,elec_num) ] BEGIN_PROVIDER [ real, elec_dist, (elec_num_8,elec_num) ]
&BEGIN_PROVIDER [ real, elec_dist_vec_x, (elec_num_8,elec_num) ] &BEGIN_PROVIDER [ real, elec_dist_vec_x, (elec_num_8,elec_num) ]
&BEGIN_PROVIDER [ real, elec_dist_vec_y, ((-simd_sp+1):elec_num_8,elec_num) ] &BEGIN_PROVIDER [ real, elec_dist_vec_y, (elec_num_8,elec_num) ]
&BEGIN_PROVIDER [ real, elec_dist_vec_z, ((-2*simd_sp+1):elec_num_8,elec_num) ] &BEGIN_PROVIDER [ real, elec_dist_vec_z, (elec_num_8,elec_num) ]
implicit none implicit none
BEGIN_DOC BEGIN_DOC
! Electron-electron distances ! Electron-electron distances

58
src/mo.irp.f
View File

@ -137,11 +137,11 @@ BEGIN_PROVIDER [ real, mo_coef_transp_present, (num_present_mos_8,ao_num_8) ]
END_PROVIDER END_PROVIDER
BEGIN_PROVIDER [ real, mo_value_transp, ((-simd_sp+1):mo_num_8,elec_num) ] BEGIN_PROVIDER [ real, mo_value_transp, (mo_num_8,elec_num) ]
&BEGIN_PROVIDER [ real, mo_grad_transp_x, ((-2*simd_sp+1):mo_num_8,elec_num) ] &BEGIN_PROVIDER [ real, mo_grad_transp_x, (mo_num_8,elec_num) ]
&BEGIN_PROVIDER [ real, mo_grad_transp_y, ((-3*simd_sp+1):mo_num_8,elec_num) ] &BEGIN_PROVIDER [ real, mo_grad_transp_y, (mo_num_8,elec_num) ]
&BEGIN_PROVIDER [ real, mo_grad_transp_z, ((-4*simd_sp+1):mo_num_8,elec_num) ] &BEGIN_PROVIDER [ real, mo_grad_transp_z, (mo_num_8,elec_num) ]
&BEGIN_PROVIDER [ real, mo_lapl_transp, ((-5*simd_sp+1):mo_num_8,elec_num) ] &BEGIN_PROVIDER [ real, mo_lapl_transp, (mo_num_8,elec_num) ]
implicit none implicit none
BEGIN_DOC BEGIN_DOC
@ -295,7 +295,7 @@ BEGIN_PROVIDER [ real, mo_value, (elec_num_8,mo_num) ]
!DIR$ VECTOR ALIGNED !DIR$ VECTOR ALIGNED
mo_value = 0. mo_value = 0.
endif endif
call transpose(mo_value_transp(1,1),mo_num_8+simd_sp,mo_value,elec_num_8,mo_num,elec_num) call transpose(mo_value_transp(1,1),mo_num_8,mo_value,elec_num_8,mo_num,elec_num)
END_PROVIDER END_PROVIDER
@ -323,12 +323,32 @@ END_PROVIDER
PROVIDE primitives_reduced PROVIDE primitives_reduced
endif endif
! Transpose x last for cache efficiency ! Transpose x last for cache efficiency
call transpose_to_dp(mo_grad_transp_y(1,1),mo_num_8+3*simd_sp,mo_grad_y(1,1),elec_num_8,mo_num,elec_num) call transpose_to_dp(mo_grad_transp_y(1,1),mo_num_8,mo_grad_y(1,1),elec_num_8,mo_num,elec_num)
call transpose_to_dp(mo_grad_transp_z(1,1),mo_num_8+4*simd_sp,mo_grad_z(1,1),elec_num_8,mo_num,elec_num) call transpose_to_dp(mo_grad_transp_z(1,1),mo_num_8,mo_grad_z(1,1),elec_num_8,mo_num,elec_num)
call transpose_to_dp(mo_grad_transp_x(1,1),mo_num_8+2*simd_sp,mo_grad_x(1,1),elec_num_8,mo_num,elec_num) call transpose_to_dp(mo_grad_transp_x(1,1),mo_num_8,mo_grad_x(1,1),elec_num_8,mo_num,elec_num)
END_PROVIDER END_PROVIDER
BEGIN_PROVIDER [ double precision, mo_lapl, (elec_num_8,mo_num) ]
implicit none
BEGIN_DOC
! Laplacians of the molecular orbitals
END_DOC
integer :: i,j
integer, save :: ifirst = 0
if (ifirst == 0) then
ifirst = 1
PROVIDE primitives_reduced
!DIR$ VECTOR ALIGNED
mo_lapl = 0.d0
endif
call transpose_to_dp(mo_lapl_transp(1,1),mo_num_8,mo_lapl,elec_num_8,mo_num,elec_num)
END_PROVIDER
BEGIN_PROVIDER [ double precision, mo_grad_lapl, (4,elec_num,mo_num) ] BEGIN_PROVIDER [ double precision, mo_grad_lapl, (4,elec_num,mo_num) ]
implicit none implicit none
BEGIN_DOC BEGIN_DOC
@ -363,24 +383,6 @@ BEGIN_PROVIDER [ double precision, mo_grad_lapl_transp, (4,mo_num,elec_num) ]
END_PROVIDER END_PROVIDER
BEGIN_PROVIDER [ double precision, mo_lapl, (elec_num_8,mo_num) ]
implicit none
BEGIN_DOC
! Laplacians of the molecular orbitals
END_DOC
integer :: i,j
integer, save :: ifirst = 0
if (ifirst == 0) then
ifirst = 1
PROVIDE primitives_reduced
!DIR$ VECTOR ALIGNED
mo_lapl = 0.d0
endif
call transpose_to_dp(mo_lapl_transp(1,1),mo_num_8+5*simd_sp,mo_lapl,elec_num_8,mo_num,elec_num)
END_PROVIDER
BEGIN_PROVIDER [ real, prepare_walkers_t ] BEGIN_PROVIDER [ real, prepare_walkers_t ]
implicit none implicit none
@ -783,7 +785,7 @@ subroutine sparse_full_mv(A,LDA, &
d14 = B4(kao) d14 = B4(kao)
d15 = B5(kao) d15 = B5(kao)
!DIR$ VECTOR ALIGNED !DIR$ VECTOR ALIGNED
do k=0,LDA-1,simd_sp do k=0,LDA-1,$IRP_ALIGN/4
!DIR$ VECTOR ALIGNED !DIR$ VECTOR ALIGNED
do j=1,$IRP_ALIGN/4 do j=1,$IRP_ALIGN/4
C1(j+k) = C1(j+k) + A(j+k,k_vec(1))*d11 C1(j+k) = C1(j+k) + A(j+k,k_vec(1))*d11