Seismic wave field three-dimensional forward modeling method based on MPI process topology

Abstract

The invention relates to a seismic wave field three-dimensional forward modeling method based on MPI process topology, and the method comprises the steps: obtaining MPI process topology parameters andthree-dimensional forward modeling parameters of all tasks, wherein each task corresponds to one shot of three-dimensional forward modeling; grouping all parallel processes based on the MPI process topology parameters, each process group comprising one or more processes, and creating a communication domain for each process group; grouping all the tasks, and allocating a group of tasks to each process group; and enabling each process group to sequentially execute each task in the corresponding task group based on the three-dimensional forward modeling parameter of each task in the corresponding task group, and output a corresponding execution result until all tasks in the corresponding task group are executed. According to the method, the three-dimensional forward modeling is carried out by adopting a multi-level fine-grained parallel optimization mode, so that the requirements of the three-dimensional forward modeling on massive memories can be met, the calculation efficiency and theexpandability are improved, and the stability of a file system is ensured.

Description

technical field [0001] The invention relates to the technical field of exploration geophysics, in particular to a three-dimensional forward modeling method of seismic wave field based on MPI process topology. Background technique [0002] In exploration geophysics, forward modeling is used to simulate the propagation of seismic waves in subsurface media, and it is the basis of techniques such as reverse time migration and waveform inversion. The 3D forward modeling is used to simulate the propagation of seismic waves in the 3D subsurface medium in the case of multiple sources (ie, multiple shots). In the case of 3D, the simulation takes a very long time and is usually performed by clusters or supercomputers. Forward modeling is the basis of seismic exploration, and its efficiency directly determines the efficiency of seismic exploration, especially for a three-dimensional work area of ​​thousands of square kilometers, the amount of calculation is very large, and the developm...

AI Technical Summary

Problems solved by technology

The single-node memory of current clusters or supercomputers is usually below 128GB, and the memory size can hardly meet the demand

If the model size is further expanded, the memory requirements are far from being met

[0005](2) Low calculation efficiency and high cost

[0006]Usual 3D forward modeling methods only carry out coarse-grained task parallelism between shots, and some will add shared memory parallel programming (Open Multi-Processing, OpenMP for short) multi-threaded parallelism, this coarse-grained processing method does not fully utilize the hardware potential of clusters or supercomputers, resulting in low computing efficiency and high cost

[0007](3) Heavy pressure on the cluster file system

[0008]Each process performs a forward simulation of one shot, reads in the 3D model and outputs the shot record, if the data volume reaches dozens of GB, hundreds of processes When I / O operations are performed at the same time, it is easy to cause excessive pressure on the file system, resulting in program crashes, and affecting the stability of the cluster or supercomputer

Images