Applied Research Of GPU In Option Pricing Calculation

Options are one kind of derivative financial instrument which the financial investors use for the conduct in the field of arbitrage and hedging transactions. Options pricing is the core issue in options trading and as well as the most complex problems in the mathematics financial applications. In the context of economic globalization, international options trading volume size and rapid growth, the current estimated option pricing platform is more and more difficult to cope with the growing data needs, and how in the shortest possible time to make a reasonable Option pricing has become a major problem plagued investors.At present, most of the option pricing calculations are run by professional software on computer platforms, without exception, which are based on central processing unit (CPU) as the core. Relative to the CPU, Graphics Processor Unit (GPU) has a better computational performance and bandwidth advantages. Especially with the recent rapid development of GPU, its applications have expanded from traditional graphics to a wider range of general-purpose computing. Thus the GPU used in financial calculations possible.Societe Generale Securities Investment Department hopes to improve its existing option pricing estimation system, as one of the topics supported by the bank, this paper chose the classical Black-Scholes option pricing model, exploring the use of GPU as a computing platform for simulation of option pricing. Author investigated how to transform the Black-Scholes equation into easy-to-algorithm model, and were prepared procedures based on the Thomas Algorithm and Crown Algorithm by using NVIDIA's CUDA technology and ported algorithm to the GPU's computing platform. By adjusting the parameters and test a lot of computation, analysis and comparison of the performance differences between GPU-based computing platforms and traditional CPU-based platform. The results show that the stability of the system to ensure the appropriate context for different parameters of magnitude of computing, GPU computing platform performance was better than CPU platform.This article as a feasibility study is to improve the existing financial option pricing system and the reasonable assumptions and the financial calculations used for the GPU provides a reasonable argument.