I found the underlying cause of the problems with --enable-fat.
./configure automatically creates fat.h with all CPU specific
functions that are available. It somehow creates entries for functions
that do not exist. If I remove all lines from fat.h that contain
either "mpn_preinv_add" or "mpn_prei
I checked through the code quickly and indeed mpz_gcdext calls
mpn_gcdext which simply implements the Lehmer algorithm (and some
optimisations for one and to limb divisions).
I'll do some work on the gcd and xgcd code for the next release. At
the present time I'm working on something different.
On Jan 12, 2:36 pm, Bill Hart wrote:
Hi,
> What Jason is saying (I think) is that currently xgcd (not gcd)
> doesn't use Moller's patches as such. It will in future, but not at
> the moment. So whatever guarantees you had before, you still have now.
> But that will soon change.
Ok, thanks fo
What Jason is saying (I think) is that currently xgcd (not gcd)
doesn't use Moller's patches as such. It will in future, but not at
the moment. So whatever guarantees you had before, you still have now.
But that will soon change.
The other issue here is that there are multiple algorithms here and