[gfortran] Front-end and library calls for large kinds

[FX Coudert]

> Wow.  I've only started thinking about this problem.

Well, I've only started thinking about this yesterday ;-)

> I've only glanced at the diff very quickly, but I
> noticed that there is no ishift10 whereas you included
> ishift16 (names could be spelled wrong here).  Is this
> intentional?

That's because ishftc is labeled with an integer kind. So the only 
possible values would be 1, 2 (for which we do a conversion, see 
trans-intrinsics.c, line 1895), 4, 8 (which were already present) and 16.

Anyway, I'd be glad if you looked at the patch. Updated version (I 
messed up with HAVE_GFC_REAL_* and HAVE_GFC_COMPLEX_*; mk-kinds-h.sh 
corrected to fix this). I added support for a few more functions in 
libgfortran.

FX
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: allkinds.diff
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20050626/d999cc2b/attachment.ksh>