https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88713
--- Comment #17 from Thomas Koenig <tkoenig at gcc dot gnu.org> ---
What an inline packing would (approximately) produce is this:
subroutine processBPP(X, BPP, N)
integer, intent(in) :: N
real, dimension(N,3), intent(out) :: X
real, dimension(N,10), intent(in) :: BPP
integer :: i
real :: tmp1(3)
real :: tmp2(6)
integer :: k1, k2
do concurrent (i = 1:N)
k1 = 0
do
if (.not. k1 < 3) exit
tmp1(k1+1) = BPP(i,k1+1)
k1 = k1 + 1
end do
k2 = 0
do
if (.not. k2 < 6) exit
tmp2(k2+1) = BPP(i,k2+5)
k2 = k2 + 1
end do
X(i,:) = fpdbacksolve(tmp1, tmp2)
end do
end subroutine processBPP
I see no timing difference for gfortran with this to the (:) version.
Chris, can you confirm this?
And is flang still faster by a factor of two if you use this version?
