http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47649
--- Comment #6 from Tobias Burnus <burnus at gcc dot gnu.org> 2011-02-08 21:18:16 UTC --- (In reply to comment #5) > I think the important hunk is: [...] > which means the ia64 __float128 support is not in a good shape, either just on > freebsd, but maybe on anything but hpux. It is unclear whether _U_Q* are > meant > to be provided by ia64 libgcc or libc. To my knowledge, libquadmath works on IA64 Linux [1-2] (which is contrary to FreeBSD GLIBC based) and also on HPUX, though there it is not used by gfortran [3]. It also works on x86-64-FreeBSD thus it looks as if only FreeBSD on IA64 is affected - maybe also other non-Linux, non-HPUX based operating systems on IA64. [1] http://gcc.gnu.org/ml/gcc-testresults/2011-02/msg00870.html [2] http://gcc.gnu.org/ml/gcc-testresults/2011-02/msg00869.html [3] http://gcc.gnu.org/ml/gcc-testresults/2011-02/msg00858.html though as on HPUX "long double" is as wide as __float128, gfortran uses calls to the long-double version for REAL(16). However, I recall that libquadmath works.
