Re: arbitrary precision
- From: Lurkos <lurkos.usenet@xxxxxxxxx>
- Date: Fri, 21 Oct 2011 12:43:44 +0200
Steven G. Kargl wrote:
GMP and MPFR are required for compilation. The internal representation
for INTEGER uses a mpz_t from GMP and the internal representation for
REAL uses a mpfr_t from MPFR. Once one compiles their Fortran code,
neither GMP not MPFR is required to run the resulting executable.
OK, thanks for the explaination.
I did
not pursue making this a permanent feature of gfortran because I did
not want to make GMP and MPFR required libraries at runtime.
I agree with you. That should not be the default.
BTW it would me useful to have an addition module to enable this
feature.
AFAIK it is already possible with new gfortran (4.6?) and libquad.
I no
longer have the patch, and I can't find it in my email archive (which may
have been a victim of hard drive failure).
Of course, the libraries should be available to a programmer, and yes,
the libraries are written in C. So, in principle, one can use ISO C Binding
to access them.
I've read a bit of MPFR source code, and it seems that its datatypes
(mpfr_t) are represented by mpfr_ptr.
Of course there is no equivalent datatype in ISO_C_BINDING and mpfr_ptr
doesn't seem to be a "struct".
Can you suggest to me how can I map it to a Fortran TYPE?
--
Lurkos
.
- Follow-Ups:
- Re: arbitrary precision
- From: Steven G. Kargl
- Re: arbitrary precision
- References:
- Re: arbitrary precision
- From: Wolfgang Kilian
- Re: arbitrary precision
- From: Lurkos
- Re: arbitrary precision
- From: Wolfgang Kilian
- Re: arbitrary precision
- From: Steven G. Kargl
- Re: arbitrary precision
- From: Lurkos
- Re: arbitrary precision
- From: Steven G. Kargl
- Re: arbitrary precision
- Prev by Date: Re: sustained number crunching
- Next by Date: Re: sustained number crunching
- Previous by thread: Re: arbitrary precision
- Next by thread: Re: arbitrary precision
- Index(es):
Relevant Pages
|