Multi-GPU-Parallel and Tile-Based Kernel Density Estimation for Large-Scale Spatial Point Pattern Analysis

Kernel density estimation (KDE) is a commonly used method for spatial point pattern analysis, but it computationally demanding when analyzing large datasets. GPU-based parallel computing has been adopted to address such computational challenges. The existing GPU-parallel KDE method, however, utilizes only one GPU computing. Additionally, assumes that the input data can be held in memory all at once computation, which unrealistic conducting analysis over geographic areas high resolution. This study develops multi-GPU-parallel and tile-based algorithm overcome these limitations. It exploits multiple GPUs speedup complex computation by distributing across GPUs, approaches with strategy bypass bottleneck. Experiment results show running on achieves significant speedups single GPU, higher are achieved tasks of larger problem size. renders feasible estimate high-resolution surfaces even limited memory. Multi-GPU estimation, while incurring very little overhead, effectively enable large-scale geospatial big data.