gmp 4.1 documentation improvement suggestions

[Nathalie Revol]

> PS: MPFI does not exactly answer this problem, since it will still compute
> with a large precision in that case.

If the operands are computed intervals which are wide (width expresses the
lack/loss of accuracy, among others), then MPFI will compute with a large
precision a wide interval (i.e. the result will have a poor accuracy).
But if the operands are given as (any precision) floating-point numbers,
Paul is right: MPFI will consider these operands as "exact" and it will
compute with a large precision a very narrow interval...

	Nathalie

---------------------------------------------------------------------------

     Nathalie REVOL                        INRIA Rhone Alpes
     LIP - Projet Arenaire                 tel : (33) 4-72-72-86-42
     Ecole Normale Superieure de Lyon      Fax : (33) 4-72-72-80-80
     69364 Lyon Cedex 07                   Nathalie.Revol@ens-lyon.fr
     FRANCE                                http://www.ens-lyon.fr/~nrevol

---------------------------------------------------------------------------