This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: [Patch, Fortran] Two minor tweaks for type-bound procedure parsing
Hi Dennis,
Dennis Wassel wrote:
2008/8/25 Tobias Burnus <burnus@net-b.de>:
[snip]
PS: g95 now started to support REAL(kind=16) on x86/x86-64 hardware [so far
only +, -, *, /].
Are there any plans to include that into gfortran anytime soon? ifort
has been supporting that for at least one minor version AFAIK, and I
remember missing it in gfortran.
Yes, there are plans, but there are only plans. And FX who thought at
looking at it, now dropped out due to lack of time.
REAL(16) needs to be done in software -- on x86, x86-64 -- as it is not
supported in hardware; if you want to use more than REAL(8) on x86,
x86-64 you can use REAL(10). Or you use a system such as PowerPC which
supports REAL(16) in silicon.
See also: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32049
Issues:
a) Everywhere is assumed that either REAL(10) or REAL(16) is supported,
but never both
b) Going beyond +-*/ (e.g. supporting cos()) requires some additional
library work (ldbl-128 in glibc?)
So much to do, so little time, I know, and I'd even start on that
myself, if it wasn't _miles_ above my head. Got enough on my plate
with my first simple patch right now :-)
Starting small, increasing project size slowly and having far reaching
dreams is ok ;-)
Still, I currently prefer to have Fortran 2003's OOP working to having a
software emulated REAL(16) - especially since I almost never even use
REAL(10).
Tobias