Batch Records Insertion into Multidimensional Linear Dynamic Hashing Table on GPU
School of Computer Science and Engineering, South China University of Technology, Guangzhou, 510006, China
Journal of Computational Information Systems 8: 10 (2012) 4293-4301, 2012
@article{liu2012batch,
title={Batch Records Insertion into Multidimensional Linear Dynamic Hashing Table on GPU},
author={LIU, Y. and XI, J. and LAI, Z. and HUANG, D.},
journal={Journal of Computational Information Systems},
volume={8},
number={10},
pages={4293–4301},
year={2012}
}
Many parallel indexing solutions of multidimensional data have been proposed on graphics processing units (GPU) platform, whereas none of them has considered the dynamic update of data. A new solution of inserting batch records into multidimensional linear dynamic hashing (MLDH) table has been presented in this paper, which has implemented lock-free batch insertion and update of hashing table by expanding the traditional data structure of hashing table and utilizing atomic functions. Comparing among the test results on various platforms, the presented solution has achieve a performance improvement of 35 times over traditional solutions of CPU batch insertion. Therefore the efficiency is greatly improved when constructing and maintaining the index of multidimensional data on GPU.
May 15, 2012 by hgpu
