arXiv subjects

Gilles Scarella

Publications and source records attributed to Gilles Scarella.

3 recordsLinked to original sources

arXiv2021

Towards Parallel CFD computation for the ADAPT framework

In order to run Computational Fluid Dynamics (CFD) codes on large scale infrastructures, parallel computing has to be used because of the computational intensive nature of the problems. In this paper we investigate the ADAPT platform where we couple flow Partial Differential Equations and a Poisson equation. This leads to a linear system which we solve using direct methods. The implementation deals with the MUMPS parallel multi-frontal direct solver and mesh partitioning methods using METIS to improve the performance of the framework. We also investigate, in this paper, how the mesh partitioning methods are able to optimize the mesh cell distribution for the ADAPT solver. The experience gained in this paper facilitates the move to the 3D version of ADAPT and the move to a Service Oriented view of ADAPT as future work.

cs.DC↗

arXiv2015

An efficient way to assemble finite element matrices in vector languages

Efficient Matlab codes in 2D and 3D have been proposed recently to assemble finite element matrices. In this paper we present simple, compact and efficient vectorized algorithms, which are variants of these codes, in arbitrary dimension, without the use of any lower level language. They can be easily implemented in many vector languages (e.g. Matlab, Octave, Python, Scilab, R, Julia, C++ with STL,...). The principle of these techniques is general, we present it for the assembly of several finite element matrices in arbitrary dimension, in the P1 finite element case. We also provide an extension of the algorithms to the case of a system of PDE's. Then we give an extension to piecewise polynomials of higher order. We compare numerically the performance of these algorithms in Matlab, Octave and Python, with that in FreeFEM++ and in a compiled language such as C. Examples show that, unlike what is commonly believed, the performance is not radically worse than that of C : in the best/worst cases, selected vector languages are respectively 2.3/3.5 and 2.9/4.1 times slower than C in the scalar and vector cases. We also present numerical results which illustrate the computational costs of these algorithms compared to standard algorithms and to other recent ones.

cs.MS↗

arXiv2013

An efficient way to perform the assembly of finite element matrices in Matlab and Octave

We describe different optimization techniques to perform the assembly of finite element matrices in Matlab and Octave, from the standard approach to recent vectorized ones, without any low level language used. We finally obtain a simple and efficient vectorized algorithm able to compete in performance with dedicated software such as FreeFEM++. The principle of this assembly algorithm is general, we present it for different matrices in the P1 finite elements case and in linear elasticity. We present numerical results which illustrate the computational costs of the different approaches

math.NA↗