Parallelization and vectorization of simulation based option pricing methods

Abstract

Simulation based pricing methods are used for a broad range of derivative valuation problems for which no closed form solution is known. They are easily adaptable to new products, and they show a superior performance for multidimensional pricing problems compared to other pricing techniques. In this paper we show how pricing methods based on Monte Carlo simulation and the stochastic mesh method of Broadie and Glasserman can be sped up by means of parallelization and vectorization. Computational results are given for multidimensional American and European pricing problems and on two different execution platforms; an MPI based NEC PC cluster and an NEC SX-6i vector computer. © Springer-Verlag 2003.