Abstract
The STAC-A2TM benchmark is an emerging standard de- signed to evaluate the speed, scalability and quality of computational platforms for performing financial risk analytics in the capital markets industry. The problem posed by the benchmark is the computation of several types of Greeks for an exotic option under an American exercise model. We recently reported record-setting performance for a STAC- A2 benchmark solution developed for an IBM® POWER8® S824 server. We explain the high performance of our solution in terms of the architecture, scalability and high memory bandwidth provided by POWER8 based systems. Developing the benchmark application also led us to investigate and perfect several techniques that are generally applicable to the simulation of complex options and their sensitivities. We describe several of these techniques in detail, along with the performance impacts we observed when compared with other approaches. We focus on two areas in particular, namely cache-efficient data management for Monte Carlo simulation of American-exercise options, and a parallel implementation of the Longstaff Schwartz algorithm.