We discuss the design and development of a parallel code for Large Eddy Simulation (LES) by exploiting libraries for sparse matrix computations. We formulate a numerical procedure for the LES of turbulent channel flows, based on an approximate projection method, in terms of linear algebra operators involving sparse matrices and vectors. Then we implement the procedure using general-purpose linear algebra libraries as building blocks. This approach allows to pursue goals such as modularity, accuracy and robustness, as well as easy and fast exploitation of parallelism, with a relatively low coding effort. The parallel LES code developed in this work, named SParC-LES (Sparse Parallel Computation-based LES), exploits two parallel libraries: PSBLAS, providing basic sparse matrix operators and Krylov solvers, and MLD2P4, providing a suite of algebraic multilevel Schwarz preconditioners. Numerical experiments, concerning the simulation by SParC-LES of a turbulent flow in a plane channel, confirm that the LES code can achieve a satisfactory parallel performance. This supports our opinion that the software design methodology used to build SParC-LES yields a very good tradeoff between the exploitation of the computational power of parallel computers and the amount of coding effort.

SParC-LES: enabling large eddy simulations with parallel sparse matrix computation tools

DENARO, Filippo Maria;DI SERAFINO, Daniela;
2015

Abstract

We discuss the design and development of a parallel code for Large Eddy Simulation (LES) by exploiting libraries for sparse matrix computations. We formulate a numerical procedure for the LES of turbulent channel flows, based on an approximate projection method, in terms of linear algebra operators involving sparse matrices and vectors. Then we implement the procedure using general-purpose linear algebra libraries as building blocks. This approach allows to pursue goals such as modularity, accuracy and robustness, as well as easy and fast exploitation of parallelism, with a relatively low coding effort. The parallel LES code developed in this work, named SParC-LES (Sparse Parallel Computation-based LES), exploits two parallel libraries: PSBLAS, providing basic sparse matrix operators and Krylov solvers, and MLD2P4, providing a suite of algebraic multilevel Schwarz preconditioners. Numerical experiments, concerning the simulation by SParC-LES of a turbulent flow in a plane channel, confirm that the LES code can achieve a satisfactory parallel performance. This supports our opinion that the software design methodology used to build SParC-LES yields a very good tradeoff between the exploitation of the computational power of parallel computers and the amount of coding effort.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11591/181739
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 9
  • ???jsp.display-item.citation.isi??? 8
social impact