Particle sorting for the projection based particle method

摘要

In this paper, we have investigated the effects of the particles sorting for projection based particle method, e.g., incompressible smoothed particle hydrodynamics. We are particularly looking at the effects of natural order sorting, Hilbert space filling curve and Morton space filling curve sorting. The data layout changes because of sorting and implementation procedure have been described in detail. We have found that the particles sorting does not only improve the general SPH kernels performance, but also has an influence on the sparse matrix structure of the pressure Poisson equation. We have also discussed and compared the performance difference between space filling curve sorting methods and natural order sorting both in serial and parallel with the dam break test case which has typical violent flow feature.