Repository navigation
Add a performance test #66
Description
Hi @jorgensd, thanks for the huge work done for the FEniCSx. As one of the many users of FEniCS/FEniCSx, I really think at least two points of FEniCSx are attracting: (1) convenience for solving PDEs and (2) the high computational efficiency based on parallel computation. I think the first point is already well illustrated in this tutorial. However, it would be better to illustrate the computational efficiency of FEniCSx versus other software such as Matlab.
Activity
jorgensd commentedon Feb 22, 2022
This is out of the scope of the tutorials. As the tutorials are hosted on binderhub, I cannot access enough processes to illustrate performance.
additionally, comparing to other software is an «infinite» task. There are a large variety of FEM software out there, written in Fortran, C++, C, Matlab, Julia etc. A tutorial cannot cover all of them, especially as many FEM software are designed in a fundamentally different way.
We have a separate code measuring performance:
https://github.com/FEniCS/performance-test
jorgensd commentedon Feb 22, 2022
@jiayingqi the results of the performance test is found at: https://fenics.github.io/performance-test-results/