Split from https://gcc.gnu.org/bugzilla/show_bug.cgi?id=34678#c47 . The test case $ cat y.f90 module y implicit none integer, parameter :: wp = selected_real_kind(15) contains subroutine foo(a,b,c) use ieee_arithmetic real(kind=wp), dimension(4), intent(out) :: a real(kind=wp), intent(in) :: b, c type (ieee_round_type), dimension(4), parameter :: mode = & [ieee_nearest, ieee_to_zero, ieee_up, ieee_down] call ieee_set_rounding_mode (mode(1)) a(1) = b + c call ieee_set_rounding_mode (mode(2)) a(2) = b + c call ieee_set_rounding_mode (mode(3)) a(3) = b + c call ieee_set_rounding_mode (mode(4)) a(4) = b + c end subroutine foo end module y program main use y real(kind=wp), dimension(4) :: a call foo(a, 0.1_wp, 0.2_wp) print *,a end program main $ gfortran -O y.f90 && ./a.out 0.30000000000000004 0.30000000000000004 0.30000000000000004 0.30000000000000004 $ gfortran y.f90 && ./a.out 0.30000000000000004 0.29999999999999999 0.30000000000000004 0.29999999999999999 shows that common subexpression removal causes the addition to be performed only once.
As long as PR 36678 is not fixed, I see one possible solution in putting a memory barrier after ieee_set_rounding_mode. This is a rather big hammer, but as long as the middle-end issue is not fixed, I do not see an alternative.
(In reply to Thomas Koenig from comment #1) > As long as PR 36678 That should be PR 34678 .
Seems to be much more complicated than I thought, see the thrad starting at https://gcc.gnu.org/pipermail/gcc-patches/2023-January/609532.html