基于MPI和OpenMP的重力及重力梯度数据并行正演算法研究

为提高重力及重力梯度数据的正演效率,笔者引入基于MPI (Message Passing Interface)和OpenMP (Open Multi-Processing)的并行计算,通过对比分析不同数据规模的网格数和模型体个数对并行效率和加速比的影响,得出随着正演数据规模的增加,并行效率和加速比均得到提高。同时对比了基于MPI和基于OpenMP的两种并行方式的性能,结果表明,重力及重力梯度数据正演的并行计算中MPI的提速能力优于OpenMP,且在较大规模数据的正演计算中基于MPI的并行效率优于基于OpenMP的并行效率。

Parallel forward modelling algorithm with gravity and gravity gradient data based on MPI and OpenMP

The parallel algorithm based on MPI(Message Passing Interface) and OpenMP(Open Multi-Processing) is introduced to enhance the forward modelling efficiency of the gravity and gravity gradient data. Through the comparison and analysis of the effects of grid number and model number of different data scale on parallel efficiency and speed-up ratio, it is concluded that the efficiency and speed-up ratio are enhanced with the increase of data scale. Meanwhile, the performance of the parallel methods based on MPI and OpenMP is compared, and the results indicate that MPI is better than OpenMP in accelerating the forward modelling parallel algorithm with gravity and gravity gradient data, and the parallel efficiency of MPI is better than OpenMP in forward modelling algorithm with large scale data.