LCPQ
/
qmcchem
mirror of https://gitlab.com/scemama/qmcchem.git
10
1
Fork 1
Code Releases Activity

Commented prefetch instructions

This commit is contained in:
Anthony Scemama 2016-11-28 15:08:55 +01:00
parent b577984de6
commit dcf19db6d6
2 changed files with 21 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
src/det.irp.f
View File

@ -1114,6 +1114,9 @@ end
endif
!DIR$ FORCEINLINE
call bitstring_to_list ( psi_det_alpha(1,det_i), mo_list_alpha_curr, l, N_int )
if (l /= elec_alpha_num) then
stop 'error in number of alpha electrons'
endif
END_PROVIDER
@ -1132,8 +1135,12 @@ END_PROVIDER
else
mo_list_beta_prev = 0
endif
!DIR$ FORCEINLINE
call bitstring_to_list ( psi_det_beta(1,det_j), mo_list_beta_curr, l, N_int )
if (l /= elec_beta_num) then
stop 'error in number of beta electrons'
endif
END_PROVIDER
BEGIN_PROVIDER [ double precision, det_alpha_value_curr ]

28
src/mo.irp.f
View File

@ -701,13 +701,13 @@ subroutine sparse_full_mv(A,LDA, &
! LDC and LDA have to be factors of simd_sp
IRP_IF NO_PREFETCH
IRP_ELSE
call MM_PREFETCH (A(j,indices(1)),3)
call MM_PREFETCH (A(j,indices(2)),3)
call MM_PREFETCH (A(j,indices(3)),3)
call MM_PREFETCH (A(j,indices(4)),3)
IRP_ENDIF
! IRP_IF NO_PREFETCH
! IRP_ELSE
! call MM_PREFETCH (A(1,indices(1)),3)
! call MM_PREFETCH (A(1,indices(2)),3)
! call MM_PREFETCH (A(1,indices(3)),3)
! call MM_PREFETCH (A(1,indices(4)),3)
! IRP_ENDIF
!DIR$ SIMD
do j=1,LDC
@ -757,13 +757,13 @@ subroutine sparse_full_mv(A,LDA, &
!DIR$ VECTOR ALIGNED
!DIR$ SIMD FIRSTPRIVATE(d11,d21,d31,d41)
do j=1,$IRP_ALIGN/4
IRP_IF NO_PREFETCH
IRP_ELSE
call MM_PREFETCH (A(j+k,indices(kao+4)),3)
call MM_PREFETCH (A(j+k,indices(kao+5)),3)
call MM_PREFETCH (A(j+k,indices(kao+6)),3)
call MM_PREFETCH (A(j+k,indices(kao+7)),3)
IRP_ENDIF
! IRP_IF NO_PREFETCH
! IRP_ELSE
! call MM_PREFETCH (A(j+k,indices(kao+4)),3)
! call MM_PREFETCH (A(j+k,indices(kao+5)),3)
! call MM_PREFETCH (A(j+k,indices(kao+6)),3)
! call MM_PREFETCH (A(j+k,indices(kao+7)),3)
! IRP_ENDIF
C1(j+k) = C1(j+k) + A(j+k,k_vec(1))*d11 + A(j+k,k_vec(2))*d21&
+ A(j+k,k_vec(3))*d31 + A(j+k,k_vec(4))*d41
enddo