Research And Development Of 3D Kriging Interpolation Algorithm Based On GPU

Geostatistics(spatial information statistics)Kriging method is a very excellent interpolation algorithm,which is widely used in the fields of Geology and mineral resources,petroleum geology and coalfield geological exploration and development.The disadvantage of this algorithm is that the calculation is complex and requires a lot of calculation,and there are obvious deficiencies in the application of high precision and strong real-time requirements.The existing research on parallel computing of Kriging method is mostly based on interpolation in two-dimensional space,but the calculation process and difficulties of three-dimensional Kriging algorithm are very different from them,and there is a lack of research on parallel computing of non parametric Kriging method and a complete solution based on GPU computing architecture.The main research contents of this paper include the development of 3D Kriging algorithm module and GPU parallel computing based on CUDA(1)In this paper,a 3D Kriging interpolation algorithm module based on Open MP and CUDA is developed according to Stanford University FORTRAN open source library GSLIB.The core sub function modules include data preprocessing module,space system module,variogram model module,neighborhood search module,CPU / GPU parallel computing module,result processing module,etc.(2)This paper proposes a CPU parallel framework based on Open MP to accelerate the 3D Kriging interpolation algorithm.The experimental results show that when applied to high-resolution interpolation calculation,the interpolation speed can be significantly improved compared with the serial algorithm(ordinary Kriging and indicative Kriging)under the same parameter configuration(specific performance is 4-core CPU,5.16 and5.09 times,8-core CPU,13.23 and 13.50 times).(3)This paper presents a new framework for Kriging interpolation on GPU.The performance of the framework is analyzed under the conditions of CPU-GPU computing platforms with different cores,interpolation resolution and parameters that have great influence on the performance of the algorithm in practical application.The experimental results show that the GPU parallel framework can achieve a maximum speedup of 60.7times for the indicator Kriging algorithm.