Monte Carlo simulation on GPGPU using prefix computation method

Random probability estimation is one of the computational intensive factors in Monte Carlo simulation. This paper presents the parallel implementation of random probability estimation for a Monte Carlo simulation. Parallel prefix computation is used to accelerate the speedup of parallel formulation of random probability estimation. The proposed work is implemented using C++ AMP (Accelerated Massive Parallelism) programming language and tested on General Purpose computation on Graphics Processing Unit (GPGPU). The experimental result shows that the average speedup achieved on GPU-based implementation is 29.61% when compared to sequential implementation of random probability estimation. The performance of the proposed work is also evaluated and compared with actual American option pricing values.