Preliminary implementation of multi-gpu communication for nonbonded force calculation; energy kernel not yet implemented