https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111655
--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Paul Eggert from comment #5)
> > The match pattern which causes the issue:
> > (simplify
> > /* signbit(x) -> 0 if x is nonnegative. */
> > (SIGNBIT tree_expr_nonnegative_p@0)
> > { integer_zero_node; })
> I don't see anything wrong with that match pattern.
>
> I speculate that what's wrong is that GCC incorrectly thinks that 0.0/0.0 is
> nonnegative. Although it's tempting to say that the sign bit of a division
> is the exclusive OR of the sign bits of its operands, evidently this is not
> true on x86-64 when NaNs are involved.
tree_expr_nonnegative_p for divide does:
case RDIV_EXPR:
case TRUNC_DIV_EXPR:
case CEIL_DIV_EXPR:
case FLOOR_DIV_EXPR:
case ROUND_DIV_EXPR:
return RECURSE (op0) && RECURSE (op1);
Since 0.0 and 0.0 both don't have their sign bits set, GCC assumes diving them
won't produce a value with the sign bit set ...
I really think x86_64 div instruction is broken for IEEE really.
