speed of mpz_odd_p and lazy allocation

[Marc Glisse]

On Sat, 18 Aug 2012, Marc Glisse wrote:

> the existence of an empty state.

Note that (as some people remarked) this empty state doesn't have to be 0, 
it makes a lot of sense for it to be NaN (in this conversation it would 
apply only to the move constructor, not the default constructor). Now 
having NaN in GMP is a different topic, I think I've sparked enough 
controversy already with this conversation ;-)

-- 
Marc Glisse