two questions on GMP & C float

Sylvain Pion Sylvain.Pion at
Sat May 16 18:42:10 CEST 2009

Torbjorn Granlund a écrit :
> Sylvain Pion <Sylvain.Pion at> writes:
>   <historical context>
>   At some point in the past, there was a goal to have mpfr_t replace mpf_t.
>   This was given up, even though some things have changed which might
>   make it worth reconsidering, like the integration of MPFR as part of
>   the GNU project.  (BTW, what's the goal of mpf_t besides staying here
>   for backward compatibility?  Does it address areas which mpfr_t does
>   not, or not nicely at all?).
>   Unfortunately the outcome of this failed merge is that there are now
>   2 multiple precision floating-point types in the GNU project, with
>   the possible confusion and interoperability issues that can arise.
>   </historical context>
> The main multiple precision fp library in the GNU project is mpfr, mpf
> is kept mainly for compatibility reasons.  It is unlikely that I will
> undertake major development projects for mpf.

Do you think the GMP documentation could state this explicitly ?
Someting like "mpf_t is mainly kept for backward compatibility,
we recommend to use MPFR instead".

> mpfr has been part of GMP for two periods, but it was determined that
> it was better to maintain the packages separately.  I don't see this as
> a bad thing.

That's OK for me too in this case.

> There are a number of extension libraries around GMP.  I think that is a
> good thing.  I do not wish to integrate them inside GMP, but more
> cooperation with these projects about useful developments in GMP would
> be good.

In general this is also fine for me.

Sylvain Pion
INRIA Sophia-Antipolis
Geometrica Project-Team

More information about the gmp-discuss mailing list