Inside frprmn code, should not it be

dlinmin(p,xi,fret,func,dfunc)

instead of

linmin(p,xi,fret,func)

It seems silly not to use the gradient function in the minimization. Or do I miss something?

Thanks in advance.

This point is discussed in the paragraph immediately before the code for frprmn. Yes, you can certainly use dlinmin. It's somewhat problem dependent whether you will see significant performance improvement, because dlinmin is more expensive than linmin.

Feel free to let us all know what you find for your problem.

Saul Teukolsky