Error of lammps + GPU using flare potential

[shihao-code]

*As I used flare (v_1.3.3) potential for lammps (22 Dec 2022) + GPU calculation (Kokkos) (i.e., lmp_mpi -k on g 1 -sf kk -package kokkos newton on neigh full -in .in > olog.txt), an error was obtained, and "nan" was output.

Step PotEng Volume Cella Cellb Cellc CellAlpha CellBeta CellGamma Press Pxx Pyy Pzz Pxy Pxz Pyz c_eatoms
0 -nan 268717.77 64.530564 64.530564 64.530564 90 90 90 -nan -nan -nan -nan -nan -nan -nan -nan

*But, if I don't used GPU with the some lammps (i.e., lmp_mpi -in .in > olog.txt), there is not error.

Step PotEng Volume Cella Cellb Cellc CellAlpha CellBeta CellGamma Press Pxx Pyy Pzz Pxy Pxz Pyz c_eatoms
0 -1707.7557 268717.77 64.530564 64.530564 64.530564 90 90 90 -9344.8769 -9344.8769 -9344.8769 -9344.8769 -2.0916633e-10 3.4853549e-10 2.1048608e-10 -1707.7557

Please see the attached log files of lammps for detail.

log.lammps_CPU.txt
log.lammps_GPU.txt

Thank you sincerely！

[YuuuXie]

Hi @shihao-code , I think your problem is related to the issue here: mir-group/flare_pp#43 (@anjohan)

(A side note, the issue is opened in the flare_pp repo, and we have merged flare_pp into flare. The current repo (flare) is the one that we are actively maintaining.)