MFEM team's opinion on the Lattice Boltzmann Method

[salazardetroya]

Hi, this is not a FEM question, but given the recent popularity of the LBM for CFD, I wanted to ask the MFEM team for their opinion. Examples of the LBM's capacity can be seen for instance in this project: https://github.com/ProjectPhysX/FluidX3D

My understanding is that the LBM is very effective for GPU chips given its local nature and structured grid. Critics of this method say that it does not represent well the physics for certain scenarios, as compared to the Navier-Stokes. They have also criticized the problems with stability and the fact that it requires many time steps for convergence. However, it seems that this can be alleviated with the power of GPUs.

So, what do you think? Is the LBM a viable alternative to FEM for CFD?