Re: Forwarded patch for modular exponentiation support (GMP powm)

[Kevin Ryde]

Eric Hanchrow <address@hidden> writes:
>
> See if the attached patch has attained sufficient grooviness.

Please direct all guile devel matters to the guile-devel list, not to
me.  If you're having trouble posting directly you might be able to go
through the gmane.org mail<->news gateway.

> And by
> the way: does the function `coerce_to_big' bother you?  It bothers me,
> because it seems necessary to me, and yet there isn't anything like it
> already.  That makes me wonder why I feel it's necessary, whereas
> everyone else got along fine without it; I figure I'm overlooking
> something.

Usually an inum is handled by separate code and one of the gmp "_ui"
functions.  And an inline mpz_t temporary when there's no _ui
function.

> +  /* if the exponent K is negative, and we simply call mpz_powm, we
> +     might well get a divide-by-zero exception.

... when an inverse 1/n mod m doesn't exist (or is not unique) ...

>       Since those are hard
> +     to handle, we'll do the inversion ourselves -- because that way
> +     we get a simple failure code, which is easy to handle. */
> +  
> +  if (-1 == mpz_sgn(k_tmp))
> +    {
> +      needs_inverting = 1;

You can do the inversion before the powm call, ie. call it with 1/n
and abs(k).  Might be a touch simpler.

> +  mpz_powm (SCM_I_BIG_MPZ (result),
> +            n_tmp,
> +            k_tmp,
> +            m_tmp);

You can use the result destination for one of the temporaries, to save
an mpz_init, if it doesn't get too messy.