Environmental modeling using graphical processing unit with CUDA

Róbert Mészáros, Ferenc Molnár, Ferenc Izsák, T. Kovács, Péter Dombovári, Ákos Steierlein, Roland Nagy, I. Lagzi

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Modeling transport and deposition processes of toxic materials in the atmosphere is one of the most challenging environmental tasks. These numerical simulations with dispersion models are very time consuming, therefore, their acceleration is extremely important. One possible, effective solution for increasing the computational time can be the parallelization of the source codes. At the same time, the technological improvement of graphics hardware created a possibility to use desktop video cards to solve numerically intensive tasks. In this study, we present a new and powerful parallel computing structure for solving different environmental model simulations using the graphics processing units (GPUs) with CUDA (compute unified device architecture). Two different types of dispersion models were developed and applied based on this technology: a stochastic Lagrangian particle model and an Eulerian model. We present and discuss the results and advantages of both methods. A Lagrangian particle model was applied to simulate the transport of radioactive pollutants from a point source after a hypothetical accidental release at local scale. In addition, an Eulerian model was used to simulate sulfur dioxide transport and transformation in the troposphere at regional scale. Moreover, in both cases, CPU and GPU computational times were also compared. We can achieve typical acceleration values in the order of 80-120 times in case of Lagrangian model and 30-40 times in case of Eulerian model using this new parallel computational framework compared to CPU using a single-threaded implementation. Next to the obvious advantages, the barriers of this new method are also discussed in this study.

Original languageEnglish
Pages (from-to)237-251
Number of pages15
JournalIdojaras
Volume116
Issue number4
Publication statusPublished - Dec 2012

Fingerprint

environmental modeling
Eulerian analysis
toxic material
parallel computing
sulfur dioxide
hardware
simulation
point source
troposphere
atmosphere
modeling

Keywords

  • Accidental release
  • Air pollution
  • CUDA
  • Environmental modeling
  • GPU computing

ASJC Scopus subject areas

  • Atmospheric Science

Cite this

Mészáros, R., Molnár, F., Izsák, F., Kovács, T., Dombovári, P., Steierlein, Á., ... Lagzi, I. (2012). Environmental modeling using graphical processing unit with CUDA. Idojaras, 116(4), 237-251.

Environmental modeling using graphical processing unit with CUDA. / Mészáros, Róbert; Molnár, Ferenc; Izsák, Ferenc; Kovács, T.; Dombovári, Péter; Steierlein, Ákos; Nagy, Roland; Lagzi, I.

In: Idojaras, Vol. 116, No. 4, 12.2012, p. 237-251.

Research output: Contribution to journalArticle

Mészáros, R, Molnár, F, Izsák, F, Kovács, T, Dombovári, P, Steierlein, Á, Nagy, R & Lagzi, I 2012, 'Environmental modeling using graphical processing unit with CUDA', Idojaras, vol. 116, no. 4, pp. 237-251.
Mészáros R, Molnár F, Izsák F, Kovács T, Dombovári P, Steierlein Á et al. Environmental modeling using graphical processing unit with CUDA. Idojaras. 2012 Dec;116(4):237-251.
Mészáros, Róbert ; Molnár, Ferenc ; Izsák, Ferenc ; Kovács, T. ; Dombovári, Péter ; Steierlein, Ákos ; Nagy, Roland ; Lagzi, I. / Environmental modeling using graphical processing unit with CUDA. In: Idojaras. 2012 ; Vol. 116, No. 4. pp. 237-251.
@article{c6312ec5c52a48cb8d8a1f1971aa7ba0,
title = "Environmental modeling using graphical processing unit with CUDA",
abstract = "Modeling transport and deposition processes of toxic materials in the atmosphere is one of the most challenging environmental tasks. These numerical simulations with dispersion models are very time consuming, therefore, their acceleration is extremely important. One possible, effective solution for increasing the computational time can be the parallelization of the source codes. At the same time, the technological improvement of graphics hardware created a possibility to use desktop video cards to solve numerically intensive tasks. In this study, we present a new and powerful parallel computing structure for solving different environmental model simulations using the graphics processing units (GPUs) with CUDA (compute unified device architecture). Two different types of dispersion models were developed and applied based on this technology: a stochastic Lagrangian particle model and an Eulerian model. We present and discuss the results and advantages of both methods. A Lagrangian particle model was applied to simulate the transport of radioactive pollutants from a point source after a hypothetical accidental release at local scale. In addition, an Eulerian model was used to simulate sulfur dioxide transport and transformation in the troposphere at regional scale. Moreover, in both cases, CPU and GPU computational times were also compared. We can achieve typical acceleration values in the order of 80-120 times in case of Lagrangian model and 30-40 times in case of Eulerian model using this new parallel computational framework compared to CPU using a single-threaded implementation. Next to the obvious advantages, the barriers of this new method are also discussed in this study.",
keywords = "Accidental release, Air pollution, CUDA, Environmental modeling, GPU computing",
author = "R{\'o}bert M{\'e}sz{\'a}ros and Ferenc Moln{\'a}r and Ferenc Izs{\'a}k and T. Kov{\'a}cs and P{\'e}ter Dombov{\'a}ri and {\'A}kos Steierlein and Roland Nagy and I. Lagzi",
year = "2012",
month = "12",
language = "English",
volume = "116",
pages = "237--251",
journal = "Idojaras",
issn = "0324-6329",
publisher = "Hungarian Meteorological Service",
number = "4",

}

TY - JOUR

T1 - Environmental modeling using graphical processing unit with CUDA

AU - Mészáros, Róbert

AU - Molnár, Ferenc

AU - Izsák, Ferenc

AU - Kovács, T.

AU - Dombovári, Péter

AU - Steierlein, Ákos

AU - Nagy, Roland

AU - Lagzi, I.

PY - 2012/12

Y1 - 2012/12

N2 - Modeling transport and deposition processes of toxic materials in the atmosphere is one of the most challenging environmental tasks. These numerical simulations with dispersion models are very time consuming, therefore, their acceleration is extremely important. One possible, effective solution for increasing the computational time can be the parallelization of the source codes. At the same time, the technological improvement of graphics hardware created a possibility to use desktop video cards to solve numerically intensive tasks. In this study, we present a new and powerful parallel computing structure for solving different environmental model simulations using the graphics processing units (GPUs) with CUDA (compute unified device architecture). Two different types of dispersion models were developed and applied based on this technology: a stochastic Lagrangian particle model and an Eulerian model. We present and discuss the results and advantages of both methods. A Lagrangian particle model was applied to simulate the transport of radioactive pollutants from a point source after a hypothetical accidental release at local scale. In addition, an Eulerian model was used to simulate sulfur dioxide transport and transformation in the troposphere at regional scale. Moreover, in both cases, CPU and GPU computational times were also compared. We can achieve typical acceleration values in the order of 80-120 times in case of Lagrangian model and 30-40 times in case of Eulerian model using this new parallel computational framework compared to CPU using a single-threaded implementation. Next to the obvious advantages, the barriers of this new method are also discussed in this study.

AB - Modeling transport and deposition processes of toxic materials in the atmosphere is one of the most challenging environmental tasks. These numerical simulations with dispersion models are very time consuming, therefore, their acceleration is extremely important. One possible, effective solution for increasing the computational time can be the parallelization of the source codes. At the same time, the technological improvement of graphics hardware created a possibility to use desktop video cards to solve numerically intensive tasks. In this study, we present a new and powerful parallel computing structure for solving different environmental model simulations using the graphics processing units (GPUs) with CUDA (compute unified device architecture). Two different types of dispersion models were developed and applied based on this technology: a stochastic Lagrangian particle model and an Eulerian model. We present and discuss the results and advantages of both methods. A Lagrangian particle model was applied to simulate the transport of radioactive pollutants from a point source after a hypothetical accidental release at local scale. In addition, an Eulerian model was used to simulate sulfur dioxide transport and transformation in the troposphere at regional scale. Moreover, in both cases, CPU and GPU computational times were also compared. We can achieve typical acceleration values in the order of 80-120 times in case of Lagrangian model and 30-40 times in case of Eulerian model using this new parallel computational framework compared to CPU using a single-threaded implementation. Next to the obvious advantages, the barriers of this new method are also discussed in this study.

KW - Accidental release

KW - Air pollution

KW - CUDA

KW - Environmental modeling

KW - GPU computing

UR - http://www.scopus.com/inward/record.url?scp=84871636077&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84871636077&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:84871636077

VL - 116

SP - 237

EP - 251

JO - Idojaras

JF - Idojaras

SN - 0324-6329

IS - 4

ER -