https://hgpu.org/?p=4276
Multi-thread implementations of the lattice Boltzmann method on non-uniform grids for CPUs and GPUs