A Parallel Access Method for Spatial Data Using GPU

Spatial access methods (SAMs) are used for information retrieval in large spatial databases. Many of the SAMs use sequential tree structures to search the result set of the spatial data which are contained in the given query region. In order to improve performance for the SAM, this paper proposes a parallel method using GPU. Since the searching process needs intensive computation but is independently examined on a lot of the MBRs of the spatial data, the spatial search function can be efficiently computed on GPU in a massive parallel way. The proposed method achieves high speed by efficiently utilizing the parallelism of the GPU throughout the whole process and by reducing the transfer latency between CPU and GPU with the memory structure which resides in GPU memory at all times and with the usage of bitwise operation for the result set. In order to measure the speedup achieved by the proposed parallel method, the execution time is compared with the sequential R*-tree that is loaded in the main memory and executed on CPU. Keywords-spatial data; GPU; CUDA;parallel computing; spatial access method