Le 27/06/2022 à 09:54, Jakub Jelinek a écrit :
Still, using __float128 when the APIs are declared for __float128 and
_Float128 when the APIs are declared for _Float128 can be better for
consistency.
I agree with that.
I was implicitly suggesting to change the libquadmath API to use the
standard syntax and types, with the assumption that it’s an internal
thing used only by fortran (on the frontend side and on the library
side). After documenting myself further, it seems that’s a wrong
assumption.
If HAVE_FLOAT128 and GFC_REAL_16_IS_FLOAT128 stand for
_Float128 and libquadmath or *f128 APIs, we'd need some macro
to tell which APIs to use, say
USE_LIBQUADMATH and USE_IEC_60559.
That would be my preference, even if we keep both __float128 and
_Float128, as it avoids exposing the detail of the provider of the math
functions in many places it’s not needed.
Half of the changes from your patch have pattern blah__float128 ||
blah_Float128 or the negation of that, and wouldn’t be needed if the
*FLOAT128 conditions spanned both __float128 and _Float128 (or
libquadmath and libc).
