I've just entered this quite interesting forum and first of all I'd like to thanks anyone who could either help me or give me any suggest.
I'm dealing with conjugate gradient, thus I implemented frpmnr alongside all the functions it needs.
I suppose the algorithm to work also in non linear case (my function cannot be approximated by any quadratic form), if it is not the case, what should I change in order to make it work?
The function is (1/tan(x)+1/tan(y)+1/tan(x+y)) where x=acos(h(g(u,v))) and g(u,v) is a NURBS and h a polynomial.The same is for y.
I checked the code several times and it seems to be correct but it finally results in an increase of the func instead of a decrease towards minimum.
Have you any ideas why?
Any way, Thanks you.

Well,since I finally managed to solve my problem, I do post the way I solved it, should someone benefit from it.
Here it is my hint.
I replaced Newton Raphson line search with Armijo line search and so far the algorithm seems to work.