Abstract
We present a sliced data structure that is effective for neighboring particle search of particle-based simulations. In the method, a grid is constructed dynamically to fit to a particle distribution. Rather than computing the grid fitting perfectly to a particle distribution where there are no voxels which does not store particle indices, we compute a grid with some margin to the distribution. This lowers the computation cost of generating the data structure. Before storing indices of particles to a grid, key values which are used to compute the index of a voxel are computed. The present data structure can be introduced to particle-based simulations on the Graphics Processing Unit (GPU) because the construction of this data structure and access to stored values can be also performed entirely on the GPU. The present data structure free us from a restriction of computation region with a fixed grid and so that we can simulate particle motions in a larger area. Moreover, the cost of the present method is low enough to be able to use for real-time applications. In this paper, we first introduce the sliced data structure and then describe how to implement it on the GPU. Then the present method is applied to particle-based simulations and quantitative evaluations are presented.
