Abstract
A parallel implementation and analysis of advanced numerical techniques to compute option prices on GPU by Abhijit Ghosh The governing equations for pricing some customized options like multi-assets options are mimicked by higher dimensional partial differential equations (PDE). Closed form solutions of these type of PDEs rarely exist in practice. In the absence of an analytic solution, advanced numerical techniques [1] are being commonly used to obtain an approximate solution of these PDEs. But solving a PDE with dimension greater than three by the well-known advanced numerical techniques take large amount of time and struggle a lot due to shortage of memory in a standard PC. To overcome the situation Monte Carlo simulations are being used for pricing options, but take ample amount of time. Solving these type of PDEs efficiently is very challenging.In recent years, general purpose computing on graphics processing unit (GPU) evolves rapidly. Some lower dimensional PDEs have been solved on GPU [2, 3]. We are interested in efficiently solving high dimensional PDEs on GPU using parallel algorithm of some advanced numerical techniques. We are also interested to analyze our approach in terms of computational speed, cost and error.