Biologically inspired algorithms applied in automatic tuning of parallel solver parameters for fast execution times

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989100%3A27740%2F24%3A10255017" target="_blank" >RIV/61989100:27740/24:10255017 - isvavai.cz</a>

Výsledek na webu

<a href="https://pubs.aip.org/aip/acp/article/3094/1/500015/3297255/Biologically-inspired-algorithms-applied-in" target="_blank" >https://pubs.aip.org/aip/acp/article/3094/1/500015/3297255/Biologically-inspired-algorithms-applied-in</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1063/5.0210563" target="_blank" >10.1063/5.0210563</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Biologically inspired algorithms applied in automatic tuning of parallel solver parameters for fast execution times

Popis výsledku v původním jazyce

In this paper, we present a way how to automatically find and set an optimal configuration for a linear system solver during a transient analysis in order to reduce computation time. Our approach is based on biologically inspired algorithms which are able to find a reasonable configuration in tens of trials, i.e., tens of timesteps in terminology of the transient analysis. This work shows a measurement within which we compared several optimization algorithms and explored their time overhead. We focus on parallel multiphysical solvers for engineering simulations developed for high-performance computing. Our experiment includes one such solver based on Finite Element Tearing and Interconnection and 4 test cases, but our solution is modular, thus, it can be connected to other similar solvers.

Název v anglickém jazyce

Biologically inspired algorithms applied in automatic tuning of parallel solver parameters for fast execution times

Popis výsledku anglicky

In this paper, we present a way how to automatically find and set an optimal configuration for a linear system solver during a transient analysis in order to reduce computation time. Our approach is based on biologically inspired algorithms which are able to find a reasonable configuration in tens of trials, i.e., tens of timesteps in terminology of the transient analysis. This work shows a measurement within which we compared several optimization algorithms and explored their time overhead. We focus on parallel multiphysical solvers for engineering simulations developed for high-performance computing. Our experiment includes one such solver based on Finite Element Tearing and Interconnection and 4 test cases, but our solution is modular, thus, it can be connected to other similar solvers.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10200 - Computer and information sciences

Projekt

—

Návaznosti

—

Rok uplatnění

2024

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

AIP Conference Proceedings. Volume 3094

ISBN

978-0-7354-4954-1

ISSN

0094-243X

e-ISSN

1551-7616

Počet stran výsledku

4

Strana od-do

'500015-1'-'500015-4'

Název nakladatele

AIP Publishing

Místo vydání

Melville

Místo konání akce

Herákleion

Datum konání akce

19. 9. 2022

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—