Fast Octree Neighborhood Search for SPH Simulations

We present a new octree-based neighborhood search method for SPH simulation. A speedup of up to 1.9x is observed in comparison state-of-the-art methods which rely on uniform grids. While our focuses maximizing performance fixed-radius simulations, we show that it can also be used scenarios where the particle support radius not constant thanks adaptive nature octree acceleration structure. Neighborhood typically consist an structure prunes space possible neighbor pairs, followed by direct distance comparisons between remaining pairs. Previous works have focused minimizing number comparisons. However, effort minimize actual computation time, find exhibit very high throughput modern CPUs. By permitting more than strictly necessary, time spent preparing and searching reduced, yielding net positive speedup. The choice structure, instead grid methods, ensures balanced computational tasks. This benefits both parallelism provides consistently intensity detailed account high-level considerations that, together with low-level decisions, enable performance-critical parts algorithm. Finally, demonstrate algorithm large-scale benchmarks experiments ratio 3 effective multi-resolution simulations.