Bit flip only function?

[David Gillies]

If I understand correctly, what Viktor is looking for is a function that
(ugly notation, sorry) performs f: mpz-> mpz, z -> z XOR
(2^mpz_sizeinbase(z,2) -1). This is surjective but not injective so f(0)
would either be undefined or, more benignly, a fixed point f(0) = f(0).


On Sun, Jul 20, 2014 at 11:23 AM, Torbjörn Granlund <tg at gmplib.org> wrote:

> Viktor Kletzhändler <vkletzhaendler at swissonline.ch> writes:
>
>   mpz_com seems to do more than flipping.
>
> We might not understand what "flipping" might mean.  mpz_com
> computes the one's complement.
>
>   For example, for a base-2 number
>   mpz_com ( 1111 1111 ) returns ( -1 0000 0000 )
>   whereas I would like _dcom ( 1111 1111 ) returns ( 0 ).
>
>   Am I missing something?
>
> So "flipping" means f(11111111) = 0, i.e., 8 consecutive ones should map
> to 0.  I assume the function "flipping" should be a bijection, so no
> other value x should have f(x) = 0.  Perhaps f(f(x)) = x should hold for
> any value x?
>
> Then, what should f(1), f(11), f(111), ... yield?  What's so special
> about 11111111?
>
> --
> Torbjörn
> Please encrypt, key id 0xC8601622
> _______________________________________________
> gmp-discuss mailing list
> gmp-discuss at gmplib.org
> https://gmplib.org/mailman/listinfo/gmp-discuss
>



-- 
David Gillies
San Jose
Costa Rica