https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81906
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |ASSIGNED
Assignee|unassigned at gcc dot gnu.org |jakub at gcc dot gnu.org
--- Comment #13 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 42809
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=42809&action=edit
gcc8-pr81906.patch
The above in patch form. Note that:
xa = fabs (operand1);
a2_52 = 2**52;
[if flag_rounding_math]
a2_52 = copysign (a2_52, operand1);
xa = xa + a2_52 - a2_52;
return copysign (xa, operand1);
is incorrect for negative operand1, because if we use negative a2_52, then we
need to use xa = operand1 + a2_52 - a2_52; rather than xa = xa + a2_52 - a2_52;
otherwise say rint (-2.5) is -2.5.
