Out of range atoms - pppm: triclinic box and user-intel

Hi! I’m having issues running triclinic simulations with pppm and the user-intel package. I’m getting out of range atoms - pppm errors before the zero timestep thermo is printed. The exact same systems are running fine (no errors) with orthogonal boxes. I checked the pppm/intel code and it looks like there is support for triclinic boxes. Are there any troubleshooting steps I can do to figure this out? Or is this an incompatibility between user-intel and triclinic boxes? I’m running the 3 March 2020 version of LAMMPS.



it should work, but we had bugs in other PPPM styles with triclinic boxes before and the pppm/intel version may have been branched off before those were fixed.
Those issues are sometimes difficult to track down. For starters, you should create a simple (e.g. just a few charged atoms/molecules) orthogonal box and then test the same input with and without switching that box to triclinic without any tilt. That should give exactly the same forces and energies for a “run 0” input.

If that still gives out-of-range errors or runs but has different forces/energies, please post the input deck as a bug report at https://github.com/lammps/lammps/issues
and one of the LAMMPS developer will have a closer look as soon as our schedule permits.



Thanks! I have posted a bug report here: https://github.com/lammps/lammps/issues/1935.