One of the major challenges in todays post-crisis finance environment is calculating
the sensitivities of complex products for hedging and risk management.
Historically, these derivatives have been determined using bump-and-revalue, but
due to the increasing magnitude of these computations does this get increasingly
dicult on available hardware. In this paper three alternative methods for evaluating
derivatives are compared: the complex-step derivative approximation, the
algorithmic forward mode and the algorithmic backward mode. These are applied
to the price of the Credit Value Adjustment for an interest rate swap, and subsequently
assessed in terms of accuracy, stability and run time. Hands-on details
are provided for the implementation along with a thorough validation framework