基于GPU的并行条件模拟算法及其在储量估算中的应用

条件模拟是一种计算非常耗时的高精度三维插值算法。针对串行条件模拟算法计算时间过长的问题,提出基于GPU的并行条件模拟算法,并进行储量估算。对条件模拟算法进行并行分析,利用GPU的高度并行性,构建CUDA通用计算开发环境,实现串行条件模拟算法到并行条件模拟算法的转换,使条件模拟算法的时间复杂度从O(n)降至O(logn)。并对西藏甲玛铜矿进行了储量估算。实验结果表明,在安装普通NVIDIA显卡的计算机以及估算精度不下降的情况下,GPU并行条件模拟的计算效率比CPU串行条件模拟的计算效率提高了60倍以上。

The parallel conditional simulation algorithm based on GPU and its application in reserve estimation

Conditional simulation is a very time-consuming algorithm of three-dimensional high-precision interpolation. To solve the time consuming problem in serial conditional simulation algorithm, we proposed a parallel conditional simulation algorithm based on GPU, and applied it to ore reserve estimation. First, this work conducted parallel analysis on the serial conditional simulation algorithm, and constructed CUDA general computing development environment based on the high parallelism of GPU. This realized the transformation of serial conditional simulation algorithm to parallel conditional simulation algorithm, and cut the time complexity of conditional simulation algorithm from O(n) to O(log n). Experimental results of the Jiama copper deposit in Tibet on a computer with NVIDIA graphics demonstrated that, compared with the serial conditional algorithm, the GPU parallel conditional simulation algorithm could improve the efficiency by more than 60 times, without estimation accuracy loss.