https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120304
--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> --- But of course then it needs to be mangled differently from long double and _Float128 too. Itanium ABI documents e for long double, g for __float128 and DF128_ for _Float128, not really sure if g isn't already taken for something else on SPARC (e.g. PowerPC had issues with that).
