https://gcc.gnu.org/bugzilla/show_bug.cgi?id=7061
--- Comment #12 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Pawel Sikora from comment #6)
> one more testcase from x86-64 / gcc-4.1:
>
> float re(float _Complex a) { return __real__ a; }
>
> is compiled to:
>
> re: movq %xmm0, -8(%rsp)
> movss -8(%rsp), %xmm0
> ret
>
> but can be optimized to `movss %xmm0, %xmm0`.
That is https://gcc.gnu.org/bugzilla/show_bug.cgi?id=48609 really.
The original testcase in comment #0 is different and be reproduced on
aarch64-linux-gnu also with -mstrict-align.
