Comparison of adaptive mesh refinement approaches on distributed computational system

The article discusses block-based and tree-based approaches to adaptive mesh refinement on regular meshes. Our goal is to compare the efficiency of approaches in solving problems on a distributed computational system. Descriptions of the approaches and features of the implementation of difference schemes using them are given. A comparison of the efficiency of the approaches in terms of speed and number of mesh elements when performing three-dimensional calculations of the development of Rayleigh–Taylor instability is made. The modeling used a predictor-corrector scheme, a Godunov-type scheme using linear interpolation, and the HLL method. Grid adaptation is applied in the area of mixing of substances. Experiments compliance of a priori estimates of the effectiveness of the approaches obtained as a result of the experiments.