https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93727
Thomas Koenig <tkoenig at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |tkoenig at gcc dot gnu.org
--- Comment #10 from Thomas Koenig <tkoenig at gcc dot gnu.org> ---
(In reply to Jerry DeLisle from comment #8)
> I have had to explore this a bit to get to a point of understanding. The
> following C example shows building the HEX float strings and the
> incantations needed. For those systems that support the 17 byte float
> format, I have not found what is needed yet.
Fortunately, there is no 17-byte format. KIND=17 a kludge to support IEEE
quad precision (which we use everywhere else) on POWER where KIND=16 is
taken up by IBM's long double format, where they put together two doubles.
That means that a number like 1.0_16 + 2.0_16**(-1000) is valid and distinct
from 1.0_16 (which violates the Fortran standard number model).
Frankly, I do not think IBM long double is something where extra effort
for implementation is warranted. Hopefully, IBM will manage the conversion
to IEEE quads (which also have hardware suport on newer POWER systems)
in time.
