Disclosed are various embodiments for GPU-based parallel indexing for concurrent spatial queries. A number of nodes in a tree to be partitioned is determined. The tree is then iteratively partitioned with the GPU. Nodes are created with the GPU. Finally, a point insertion is performed using the GPU.USF inventors have developed a construction tree algorithm in G-PICS. Experimental results show performance boosted up to 50X, in both throughput and query response times, over best known parallel GPU and parallel CPU-based spatial query processing systems. Furthermore, the G-PICS design can be easily extended to index datasets that are too large to be placed in GPU global memory. Applications such as geographic information systems (GIS), mobile computing, scientific computing, epidemic simulation, astrophysics may be benefited by this algorithm.
Brochure
