Eventos Anais de eventos
- Você está em:
- Home
- Eventos
- Anais de eventos
- Acesso livre
EPTT 2022
13th Spring School on Transition and Turbulence
Lid-Driven Cavity simulation using the Chapel programming language
Submission Author:
Anna Caroline Felix Santos de Jesus , ES
Co-Authors:
Anna Caroline Felix Santos de Jesus, Livia S. Freire, Nelson Dias
Presenter: Anna Caroline Felix Santos de Jesus
doi://10.26678/ABCM.EPTT2022.EPT22-0060
Abstract
Most softwares in the context of computational fluid dynamics are developed with the languages C or Fortran, due to their low computational cost. These languages, however, require additional tools such as MPI or OpenMP for parallelism. In this context, the performance of the new programming language Chapel was investigated. This language purports to be fast like Fortran, portable like C and easy to code as Matlab, with a parallelization that is significantly simpler to implement compared to MPI. A code was implemented to solve the two-dimensional, transient, incompressible lid-driven cavity flow in the Chapel 1.24 and Fortran languages. The implementation is based on the finite volume method, incorporating the classical projection method to decouple the velocity and pressure fields. Preliminary verification results using the method of the manufactured solution demonstrate a correct implementation and a convergence of second order in space. Moreover, the velocity profiles obtained for Reynolds number 100 with different meshes are in agreement with literature data. Regarding the performance of the new language, in serial, Chapel 1.24 compiled with the --fast flag presented a higher computational cost compared to Fortran using the GNU compiler and the -Ofast optimization flag. However, the difference in computational cost between the languages decreased with increase in the number of grid points, indicating a tendency of similar cost for heavier simulations. Finally, the parallel version of the Chapel code was obtained by including simple commands, resulting in a low speedup that also improved with increase in the number of grid points. For a better evaluation, future work will test heavier simulations in a cluster.
Keywords
Computational fluid dynamics (CFD), Incompressible flows, Finite volume method, Projection Method, Chapel, Fortran