A Special Sorting Method for Neighbor Search Procedure in Smoothed Particle Hydrodynamics on GPUs

Nowadays, the GPGPU technology has accelerated many physics-based simulations. SPH (Smoothed Particle Hydrodynamics) is widely used to simulate complex physical phenomena, which using neighbor search as its key procedure. Sorting and grid based method is one of the widely used neighbor search methods on the GPU. This paper presents a new special sorting method to accelerate the neighbor search procedure for SPH systems on the GPU. With our method, the sorting procedure will use the result data of the previous simulation step. Compared with the standard thrust implementation, our method can accelerate the whole simulation when the scale is not very large (nearly 2.3x speedups with 8192 particles in 16*16*16 grid size). While the scale is large, it also does not reduce the performance of the whole simulation.