https://hgpu.org/?p=8378
Platform-independent parallelization of the Lattice Boltzmann method with OpenCL