> Then you could redefine f(p,q) over all of C as follows > > fnew(p,q)=f(p,q)+prod(b-A*q)
Thanks for this suggestion. Unfortunately, C is not easy to write out in that form. It would have a very large number of vertices which, with dimensions > 2, would lead to far too vast an "A" term for it to be computationally feasible.
C can be described in terms of an optimization process (e.g. I give it "p", the slopes of a hyperplane, and it is easy to determine the corresponding point(s) on C).
Sorry for not being explicit about this earlier, but I was trying to keep my post short.