Gravity gradient data sensitivity matrix compression and calling method based on multiple GPUs

Abstract

The invention provides a gravity gradient data sensitivity matrix compression and calling method based on multiple GPUs, and relates to the technical field of geophysical forward and reverse modeling. The method comprises: firstly, establishing an equivalent calculation formula of a sensitivity matrix of each component of gravity gradient data, then calculating parameters such as the range and the number of sensitivity matrix elements divided in each GPU, the compressed sensitivity matrix being subjected to parallel calculation by utilizing the GPU, and finally obtaining an index of the sensitivity matrix elements to participate in calculation, and converting the index in the compression matrix and reading a numerical value so as to realize rapid forward and reverse modeling of different gradient components or full-tensor gradient data of the large-scale gravity gradient data. Through verification, the method can effectively improve the calculation scale of gravity gradient data forward and reverse and the calculation efficiency of large-scale gravity gradient data joint inversion; the method is also suitable for the forward and reverse modeling field of other types of potential field data such as gravity and magnetic anomalies, and is also suitable for the situation of a single GPU.

Description

technical field [0001] The invention relates to the technical field of geophysical forward and inversion, in particular to a multi-GPU-based gravity gradient data sensitivity matrix compression and call method. Background technique [0002] Gravity gradient data is a kind of high-precision geophysical data with higher signal-to-noise ratio. Compared with the application of gravity anomaly, the full tensor gradient data contains more geological information, and the joint inversion of multi-component gradient data can obtain a higher resolution three-dimensional density distribution of underground space. The forward and inversion of potential field data requires the calculation of the sensitivity matrix. However, a small amount of observation data and the number of underground grid layers can also generate a large-scale sensitivity matrix, which seriously reduces the efficiency of forward and inversion calculations. At the same time, it takes up a lot of memory, resulting in ...

AI Technical Summary

Problems solved by technology

However, limited by hardware attributes such as bandwidth, the current method of using a single GPU will encounter situations such as increased communication overhead and reduced computational efficiency when processing complex iterative inversions; while using multiple GPUs can reduce the data throughput of each GPU, Improve algorithm parallelism

Images