Performance enhancements for USER-DPD 1) For an ideal fluid benchmark: ~15% speedup and ~5% memory savings 2) Removes duCond and duMech, and replaces dvSSA with v_t0 (ghosts only) 3) Eliminates the need to transmit deltas and uCond/uMech in forward_comm 4) Reduces adds and loads for computing updated velocities from v + dvSSA