http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51616
--- Comment #7 from John David Anglin <danglin at gcc dot gnu.org> 2011-12-30 15:40:59 UTC --- (In reply to comment #5) > hpux11 appears to be yet another OS that has sufficient quad > support that gfortran detects a REAL(16) type, but the OS > lacks the basic libm functions. Except possibly for hpux11.31, the OS only provides support for basic arithmetic operations. These are soft implementations. GCC has supported these operations for many years. This is why gfortran detects a REAL(16) type. The support is sufficient to build libquadmath. As far as I know, it works correctly (no errors in testsuite). There was some discussion last year about adding support to provide `l' aliases to the `q' routines in libquadmath.
