issues/limitations with large arbitrary floating point calculations and GMP

Kevin Ryde user42 at
Sat Dec 13 08:40:13 CET 2003

Klaus Berg <klaus_berg96 at> writes:
> Also, if I write software that uses this library, is it true that it
> may run differently on different hardware architectures/classes of
> machines ?  In other words, the same floating point calculation on 1
> machine could have a different result than another machine would
> have.

Yep.  Currently precision is rounded up to a multiple of a limb, so on
a 64-bit machine you might get a little more than on a 32-bit machine.

MPFR ( will give exact-bit rounding, if you want that.

More information about the gmp-discuss mailing list