[gmx-developers] Use of pbc_dx in bonded calculations

[Vishal Vaidyanathan]

Hi,
   I'm working on gmx 3.1.4 and I see that the bonded functions use
pbc_rvec_sub() to calculate the bond vectors. Is this necessary? It seems
that shift_self() is called to make molecules whole before the bonded
functions in do_force(), so a simple vector difference should give the
same result as pbc_dx() right?

   If this is not true, could somebody let me know what exactly
shift_self() does?

   Any help is appreciated!
   Thanks,
Vishal