Home

Welcome to the CUDA-MultiDimensionalIndexing wiki!

Recent research has shown promising results on using graphics processing units (GPUs) to speed up database processing. In this paper, we explore GPU-based multidimensional access methods for in-memory static data. We first develop a GPU-based access framework consisting of guidelines and primitives. The guidelines are oriented to the GPU’s hardware characteristics such as massive multithreading, device memory and on-chip stores. The data-parallel primitives are building blocks for general bottom-up or top-down bulk loadings and general single- or multiple- path searches. Based on this framework, we carry out the access methods for three representative indexes, namely the grid file, the quadtree and the R-tree. We propose or adopt necessary structural variants, such as the hierarchical grid file and the Hilbert-packed R-tree. We have implemented these methods on a GeForce 8800 GPU, and also implemented the optimized multithreaded counterparts on a Core 2 Duo CPU. The GPU algorithms are generally several times faster than the CPU counterparts. Index