https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92597
--- Comment #16 from Klaus Leppkes <leppkes at stce dot rwth-aachen.de> ---
So after reading a bit more, from my understanding, +g should be used for all
integral/floating point/vector types and +m for other (aggregate types) which
cannot end up in an register.
Still, it might be possible for an aggregate type like
struct mycomplex {
double v1;
double v2;
};
to end up in SSE register, right?
