Re: arbitrary precision

*Wolfgang Kilian* wrote:

Don't forget the GNU projects: GMP/MPFR/MPC

That's right and probably the better solution. However I couldn't find a
Fortran interface. Are you aware of some projects?

These are C libraries, but since they are requirements of the gcc
compiler suite, they may actually be automatically available when you
are using gfortran (some gfortran expert may correct me).

program precreal
implicit none

integer,parameter :: p4 = selected_real_kind(p=6)
integer,parameter :: p8 = selected_real_kind(p=15)
integer,parameter :: p10 = selected_real_kind(p=18)
integer,parameter :: p16 = selected_real_kind(p=33)
integer,parameter :: p32 = selected_real_kind(p=66)

print *, p4
print *, p8
print *, p10
print *, p16
print *, p32

endprogram precreal

Output with gfortran 4.6:
4
8
10
16
-1

It seems not possible to go beyond quadruple precision in a simple way.

By the way, what is the standard return value of HUGE, EPSILON etc. if
the KIND happens to be arbitrary-digits/arbitrary-precision ...? (I
guess that's just theory right now)

That is interesting. Actually I think this is dependent on how much
digits you take, but I don't know if they are implemented also in
arbitrary precision libraries.

--
Lurkos
.