GPU Implementation of the Finite Element Method

Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21340%2F14%3A00212999" target="_blank" >RIV/68407700:21340/14:00212999 - isvavai.cz</a>
Result on the web
—
DOI - Digital Object Identifier
—

Result language
angličtina
Original language name
GPU Implementation of the Finite Element Method
Original language description
This paper investigates a way of assembling the system matrix for nonlinear evolution problems discretized by the finite element method entirely on the GPU to avoid expensive copying of data between CPU and GPU memory. We present a general CUDA implementation of the finite element matrix assembly relying on a coloring of the computational mesh. There are no particular limitations on the mesh, so it can be unstructured. We validate the CUDA implementation by comparison with a corresponding CPU implementation. It can be seen in the paper that the GPU implementation performed 3–10 times better than the OpenMP CPU implementation and up to 25 times better than the sequential CPU implementation.
Czech name
—
Czech description
—

Project
<a href="/en/project/TA01020871" target="_blank" >TA01020871: ADVANCED CONTROL AND OPTIMIZATION OF BIOFUELS CO-FIRING FOR POWER AND ENERGY</a><br>
Continuities
P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Publication year
2014
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection
Seminar on Numerical Analysis & Winter School: Proceedings of the Conference SNA '14
ISBN
978-80-87136-16-4
ISSN
—
e-ISSN
—
Number of pages
3
Pages from-to
11-13
Publisher name
Ústav Informatiky AV ČR, v.v.i.
Place of publication
Praha
Event location
Nymburk
Event date
Jan 27, 2014
Type of event by nationality
WRD - Celosvětová akce
UT code for WoS article
—

Similar results(10)