Robert Dodier wrote:
>
>> 2) If each of the variables used as argument to the lbfgs call is a bfloat
>> object, so are the functions employed in the fit?
>>
>
> lbfgs uses only ordinary floating point arithmetic.
> It can't use Maxima's bigfloat arithmetic, sorry.
>
>
Sometime back, as a gross proof-of-concept, I took the f2cl-generated
BLAS code and ran a program over that and converted all the calls to
arithmetic functions to generic arithmetic functions. The result was a
bfloat version of BLAS that appeared to work.
Now that we have a bigfloat class to make bigfloat arithmetic easier to
use from Lisp, this could be revisited again. Perhaps not the best
solution, but it's a quick way to take proven code and make it support
bigfloats. Of course, any hardwired numbers in the algorithms will now
be incorrect, so that will have to be fixed.
Anyway, that's one option for bigfloat versions of some algorithms.
Ray