Below are some results for well-known schemes. Predictor Corrector and Crank Nicolson give good results.
Milstein and Euletr methods give less accurate results in this case.
We use a class hierarchy of SDEs for 1 and n factors using a combination of inheritance and templates.
Extra functionality is achieved with _Visitor_ and _Decorator_. We just create new derived classes and can quickly test new models and schemes.