The aim of this work is to assess the performance of a Large Eddy Simulation (LES) in-house code for compressible flows on a 2064 cores Linux computing system at the Inter-University Consortium for the Application of Super-Computing for Universities and Research (CASPUR-Rome-Italy). The code is written in Fortran90 using the Message Passing Interface (MPI) parallelization library. A non-dissipative sixth-order finite difference scheme is employed in order to solve the Favre-filtered governing equations, with a low-pass sixth order spatial filtering scheme to avoid the growth of high frequency modes, and a fourth order Runge-Kutta scheme to advance the solution in time. The open Navier–Stokes characteristic boundary conditions are implemented to avoid reflection of pressure waves and the Smagorinsky model is used to compute the subgrid scale turbulent viscosity. The compact-Padè tridiagonal system of equations is solved by employing an efficient parallel partition LU algorithm. In order to assess the performance of the parallelized structure of the code, its speed-up has been evaluated versus the linear speed-up. The model has been validated by performing simulations of an n-heptane non-reacting plane jet issuing into air at high pressure and temperature.
High-Performance Computations of Large Eddy Simulations of Compressible Jets
VIGGIANO, ANNARITA;MAGI, Vinicio
2010-01-01
Abstract
The aim of this work is to assess the performance of a Large Eddy Simulation (LES) in-house code for compressible flows on a 2064 cores Linux computing system at the Inter-University Consortium for the Application of Super-Computing for Universities and Research (CASPUR-Rome-Italy). The code is written in Fortran90 using the Message Passing Interface (MPI) parallelization library. A non-dissipative sixth-order finite difference scheme is employed in order to solve the Favre-filtered governing equations, with a low-pass sixth order spatial filtering scheme to avoid the growth of high frequency modes, and a fourth order Runge-Kutta scheme to advance the solution in time. The open Navier–Stokes characteristic boundary conditions are implemented to avoid reflection of pressure waves and the Smagorinsky model is used to compute the subgrid scale turbulent viscosity. The compact-Padè tridiagonal system of equations is solved by employing an efficient parallel partition LU algorithm. In order to assess the performance of the parallelized structure of the code, its speed-up has been evaluated versus the linear speed-up. The model has been validated by performing simulations of an n-heptane non-reacting plane jet issuing into air at high pressure and temperature.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
