refactor diffusion coeff. computing
Rigth now, it's dispatched over many kernel. This is both confusing and inefficient (although not a bottleneck).
Rigth now, it's dispatched over many kernel. This is both confusing and inefficient (although not a bottleneck).