[gmx-users] The fftw and mpi

[Tanping Li]

Dear all,

I simulate the dynamics of the protein solution in the
italium cluster by 8 nodes. I use pme. The problem is:
when I use --mdrun rerun-- function to calculate
energy with different nodes of tpr file, gromacs gives
me different Coul_LR, while other energy terms do not
change. Can I believe that gromcs give me wrong
Coul_LR in parallel?

Does the problem come from the mismatch between the
mpi and fftw? I know little about this issue. Is there
anyone who has the same experience? Thank you very
very much. I am totally stuck now.


Tanping