i have something to ask
i read the code about neighbour searching, i think the algorithm is the so called verlet list, in this method, a real neighbour list is generated. and my question is why verlet list not cell linked list (not a real neightbour list, but a neighbour cell list with reordered particles), as i know, CLL is the better considering the balance of memory usage and efficiency. Moreover, i want to implement the adaptive particles refinement base on GPUSPH (multi-gpu). I think in this situation, CLL is better than VL considering the multi-resolutions.
as i said above, i want to implement APR, i have done it before, but it is based on single GPU, so actually i have no experimence about multiGPU, and any suggestion would be wonderfui.
some state-like variation like state, policy, dirty. i dont know how it work, and what they means.