最小二乘蒙特卡洛美式期权定价的GPU实现

Implementation of the least squares Monte-Carlo American option pricing on GPU

蒙特卡洛模拟法常用来进行期权定价,但此算法存在运算量过大的问题.利用图形处理器(GPU)超强计算能力实现美式期权定价,在GPU上,首先优化实现了均匀随机数生成器,然后利用Box-Muller随机数转换算法产生随机数,最后优化实现了最小二乘蒙特卡洛模拟法的美式期权模拟定价系统.测试结果表明,GPU实现的最小二乘蒙特卡洛美式期权定价对比CPU的实现加速比最高达到了16.1.利用GPU的编程技术以更小的硬件代价,更高的执行效率,更好地完成由CPU完成的传统任务,较好地解决了蒙特卡洛模拟法运算量过大的问题,充分挖掘了GPU的通用计算潜力.

More often than not Monte-Carlo is connected with option pricing,but it is burdened with calculations.While GPU,with its stupendous calculating ability,can realize American option pricing.First of all,Mersenne twister uniform random number generator is optimized;secondly,Box-Muller conversion algorithm is optimized to realize the random number;lastly,American options pricing simulation system of the Least Squares Monte-Carlo is optimized as well as.The test indicates that the speedup ratio,GPU implementation of the least squares Monte-Carlo American option pricing is 16.1respectively,compared to CPU.Thus programming technique of GPU can be used to deal with the traditional tasks more effectively,making use of its calculating advantage instead of facing the overmuch calculating problems like Monte-Carlo.