Re: GMP bignum results using double cells.

[Kevin Ryde]

Rob Browning <address@hidden> writes:
>
> The fact that all the values are so close makes me wonder if much of
> the cost is just interpreter related.

Perhaps bigger numbers would show up better.  The advantages of gmp
normally increase with increasing size.

> Note that I left the same algorithm as before for the inum case.

mpn_gcd_1 would be a possibility for that, to save some code.
Probably no great difference in speed normally, though gmp does have
some nice assembler versions for K6 and Athlon.

>     } else if (SCM_BIGP (y)) {
>       SCM result = scm_i_mkbig ();
>       SCM mx = scm_i_mkbig ();
>       mpz_set_si(SCM_I_BIG_MPZ (mx), SCM_INUM (x));
>       scm_remember_upto_here_1 (x);
>       mpz_gcd(SCM_I_BIG_MPZ (result), SCM_I_BIG_MPZ (mx), SCM_I_BIG_MPZ (y));

Could probably use mpz_gcd_ui instead of converting x to an mpz.

Unless I've missed something basic and a conversion should be done.
In which case maybe the result variable could be pressed into service,
convert into that and call mpz_gcd(result,result,y).