Re: [Dclib-devel] Root-finding algorithm in DLib

[Davis K. <dav...@us...>]

You mean copied dlib's code and changed double to long double?  That would
probably work fine.

I'm using double in the optimizers because it's accurate enough for nearly
everything.  You can use an objective function that internally uses long
double or whatever you want, but the optimizers themselves run at double
precision.  Are you saying that you want to run the optimization until it
gets a solution that is accurate to more than about 14 digits of
precision?  If not then double is fine.  If so, then what are you doing
where you need 15 or more digits of precision in the solution?

Cheers,
Davis