Přístupnostní navigace
E-application
Search Search Close
Publication detail
MAŠEK, J. VOŘECHOVSKÝ, M.
Original Title
Parallel Implementation of Dynamical Particle System for CUDA
Type
conference paper
Language
English
Original Abstract
The presented paper deals with possible approaches to implementation of solution of dynamical particle system on the NVIDIA CUDA C/C++ platform, namely in scenarios of using the NVIDIA graphic cards with cores of the Kepler family. A simulation of a particle system is typically a computationally demanding task. However, it is possible and beneficial to make the computer implementation of the solution as parallel as possible. This degree of parallelism of the code is limited mainly by capabilities of the hardware used and by the requirements on numerical accuracy. The presented contribution discusses methodology of the solution algorithm, its parallelized CUDA C performance as well as a derivation and reasoning of correct potential formulation with respect to number of dimensions Nvar and points Nsim as the originally proposed Audze-Eglajs potential has been found to be malfunctioning.
Keywords
Particle dynamical system; parallel implementation; CUDA; serialization of threads; atomic operations; looping.
Authors
MAŠEK, J.; VOŘECHOVSKÝ, M.
Released
30. 5. 2017
Publisher
Civil-Comp Press
Location
Stirlingshire, UK
ISBN
1759-3433
Periodical
Civil-Comp Proceedings
State
United Kingdom of Great Britain and Northern Ireland
Pages from
1
Pages to
12
Pages count
URL
http://www.ctresources.info/ccp/paper.html?id=9256
BibTex
@inproceedings{BUT136620, author="Jan {Mašek} and Miroslav {Vořechovský}", title="Parallel Implementation of Dynamical Particle System for CUDA", booktitle="PROCEEDINGS OF THE FIFTH INTERNATIONAL CONFERENCE ON PARALLEL, DISTRIBUTED, GRID AND CLOUD COMPUTING FOR ENGINEERING", year="2017", journal="Civil-Comp Proceedings", pages="1--12", publisher="Civil-Comp Press", address="Stirlingshire, UK", issn="1759-3433", url="http://www.ctresources.info/ccp/paper.html?id=9256" }