Multicore and Manycore Algorithms for Octrees
Department of Computer Science and Engineering, Indian Institute of Technology, Bombay, Mumbai
Indian Institute of Technology, 2012
@phdthesis{saida2012multicore,
title={Multicore and Manycore Algorithms for Octrees},
author={Saida, D.},
school={Indian Institute of Technology, Bombay},
year={2012}
}
Octrees and compressed octrees are frequently used to represent data in an hierarchical form for high performance computing, graphics and database applications. Applications like N-body problems require building octrees multiple times. Therefore, efficient construction of octrees is critical to the efficiency of the entire applications. With ever increasing data size, there is a requirement to develop parallel algorithms for creating large octrees. We propose efficient algorithms to create octrees in parallel for manycore and multicore systems which use shared memory architecture and also give their runtime analysis. Our experimental results show that GPU octree implementation is 11.85 times faster than its sequential implementation on CPU.
July 17, 2012 by hgpu
