Exporting a gmp_rndstate_t?

[Pedro Gimeno Fortea]

On 06/06/2007 11:05:18 AM, Christopher Creutzig wrote:

>  is it possible to (portably) export and re-import a gmp_rndstate_t,
> as an mpz_t can be via mpz_export and mpz_import (plus storing the  
> sign)? (If this limitation is necessary, an answer only valid for the  
> MT algorithm would probably be fine.)

It is in the plans, just not yet done. The current version does not  
support exporting the state and there's no patch submitted yet.

>  On a slightly related note, I'd be happy to find a way of converting
> a gmp_rndstate_t to an (opaque) mpz_t which can be converted back to  
> a gmp_rndstate_t without the runtime overhead required for the  
> current initialization from an unsigned int. Here, I don't really  
> bother whether the random state remains the same, I just need a  
> deterministic outcome.

It's also in the plans to speed up the seeding process. A patch for  
that is already submitted and hopefully the next version will include  
it.

Anyway you're asking for a way to export to mpz_t; do you really need  
that format? I mean, the plans are to include an export routine which  
stores the result in an opaque structure with the user-supplied address  
and with a library-supplied length, plus a complementary import routine  
which accepts the result from the export one. I think this should be  
enough for all purposes. It puzzles me why you want to use an mpz_t  
integer for that; that specific requirement is not in the plans.

Would you please clarify that point?

-- Pedro Gimeno