This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: [RFC] Generalizing the library to arbitrary floating-point modes
- From: Richard Henderson <rth at twiddle dot net>
- To: François-Xavier Coudert <fxcoudert at gmail dot com>
- Cc: Paul Thomas <paulthomas2 at wanadoo dot fr>, Tobias dot Schlueter at physik dot uni-muenchen dot de, gfortran <fortran at gcc dot gnu dot org>
- Date: Wed, 7 Dec 2005 09:00:37 -0800
- Subject: Re: [RFC] Generalizing the library to arbitrary floating-point modes
- References: <19c433eb0511300536u5891b76j114c13a5f73ff6ef@mail.gmail.com> <1133359319.438db0d799399@www.cip.physik.uni-muenchen.de> <438DD573.2050301@wanadoo.fr> <19c433eb0511300843l3ac14e4by4ba9e6faca5471cd@mail.gmail.com> <20051201020047.GA4983@twiddle.net> <19c433eb0512070436v31844914qdfe682513323e0c5@mail.gmail.com>
On Wed, Dec 07, 2005 at 01:36:23PM +0100, François-Xavier Coudert wrote:
> The sin() function is a generic name, and it is up to the fortran compiler
> to compile it into a call to, for example, the libm sinl() function. But how
> does the front-end know that real(kind=10) corresponds to the l-suffixed
> math functions, and not (as is the case on ia64-hpux, where it corresponds
> to sinw) to something else. Currently, this is hardcoded, but that's just a
> short-term hack around the real problem.
It isn't a short-term hack. The Fortran front end has access to the
exact same backend configuration macro that the C front end uses to
select types. So it really does know with absolute certainty what
kind "long double" corresponds to.
> There is other information that the front-end would need, that it currently
> doesn't. For example, for some I/O function calls, we need the ability to
> create function calls with off_t arguments.
Also solvable.
r~