Variable precision distance search for random fractal cluster simulations

  • Authors:
  • Sosa-Herrera Antonio;Rodriguez-Romo Suemi

  • Affiliations:
  • Universidad Nacional Autónoma de México, CuautitlánIzcalli, Edo. Mex., Mexico;Center of Theoretical Research, Universidad Nacional Autónoma de México, CuautitlánIzcalli, Edo. Mex., Mexico

  • Venue:
  • WSEAS Transactions on Computers
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

A simple an effective algorithm for performing distance queries between a large number of points stored in quadtrees and octrees. The algorithm is developed and tested for the construction of diffusion-limited aggregates. To achieve an enhancement on the searching time we accept approximate distance values with low precision at the first levels of the hierarchical structure, and accurate ones at the last level. The structure of the trees is the only feature used for the determination of approximate distances at any stage. These techniques allowed us to build DLA clusters with up to 109 particles for the two-dimensional case and up to 108 particles for the three-dimensional case. We also worked with the PDLA model obtaining fractal clusters with up to 1010 and 109 particles for two and three dimension clusters respectively. We worked on a supercomputer to run the PDLA simulations, as well as a high performance server for DLA simulations. We employed POSIX threads to provide parallelization and mutexes as control mechanisms to achieve synchronization between groups of 4 processors, hence simulating PDLA clusters with early convergence to the DLA model.