【摘要】 提出了一种在CNGrid网格服务环境下解决期权定价问题的并行应用方法.这种方法基于BSDE(backward stochastic differential equation)模型.根据异构计算资源的特点,使用CUDA和MPI分别在GPU计算节点和CPU计算节点上实现并行算法,比较不同编程在异构计算节点上的实现效率.通过监控计算节点上计算任务的负载状况,利用CNGrid所提供的计算服务,灵活地在异构计算节点上完成期权定价计算任务.更多还原

【Abstract】 A parallel application for option pricing in CNGrid was presented.The paralleling method was based on BSDE(backward stochastic differential equation) model.It can improve the accuracy and effectiveness of option pricing.According to the characteristics of the heterogeneous resources in CNGrid,CUDA(compute unified device architecture) and MPI(message passing interface) to implement the parallel algorithm on GPU and CPU nodes respectively.The performance of parallel algorithm on the CNGrid platform that supports different architectures was evaluated.Monitoring the system load conditions on compute nodes,the financial application services provided by CNGrid was used to implement the option pricing flexibly.更多还原