mpf exponent range

Torbjorn Granlund tg at
Tue Aug 23 07:26:20 CEST 2011

Zimmermann Paul <Paul.Zimmermann at> writes:

  > Do people on this list have any opinions about
  > the exponent range, of mpf, in particular, if we should make input and
  > output work for the entire allowed number range?
  my opinion is yes. If input/output functions do not work for the entire
  allowed number range, this should be documented, so that the user knows
  what are the exact limits.
I agree that it should at least be documented.

  By the way the documentation says:
        The exponent of each float is a fixed precision, one machine word on
     most systems.  In the current implementation the exponent is a count of
     limbs, so for example on a 32-bit system this means a range of roughly
     2^-68719476768 to 2^68719476736, or on a 64-bit system this will be
  It would be nice to make precise the bounds on a 64-bit system. I guess
  the maximum is 2^(2^63*64), but what is the minimum?
I'd expect 2^-(63*64+64).


More information about the gmp-discuss mailing list