MD neighbor lists

Hi. I am currently working on parallel molecular dynamics simulations
 of molecular systems. I have tried various methods of forming the neighbors
 list, such as linked-cells and the standard
 Verlet neighbors. Does anyone have a way of making the neighbor list
 that they feel is particularly good for parallelism that they would
 like to share? I haven't found any scheme that seems to be significantly
 better than the others. Also, the neighbor list has to be made with the
 exclusion of atoms that are 1-2 or 1-3 (maybe 1-4) bonded. Thanks in
 advance for any suggestions. Please mail me directly because "reply"
 is not working well at our site. John Nicholas
 nicholas { *at * }