
Barry Smith authored
If one is updating the model parameters (and hence changing the f(x) at each Newton step (with SNESSetUpdate())) the norm of f used by the line search must be updated or otherwise the line search will use the previous fnorm which may be very small and so no sufficient reduction can take place. The extra Norm is a noop unless the vector has been changed by the user since vectors cache their norms Reportedby: Teferra, Kirubel
3238a83e
Loading