Hi.
I haven’t got time to fully check it out, but a 2 processor run of 1Feb14 revealed that it was stuck in two different MPI_Allreduce calls:
group.cpp line 844 (the xcm routine)
dump write (line 302)
It look like both went into verlet->modify (line 144 of verlet.cpp) and one came out whilst the other got stuck in xcm.
Hope that helps,
Duncan
From: Axel Kohlmeyer [email protected] - 2014-04-06 08:15:56
|
Hi.
I haven't got time to fully check it out, but a 2 processor run of 1Feb14
revealed that it was stuck in two different MPI_Allreduce calls:group.cpp line 844 (the xcm routine)
dump write (line 302)It look like both went into verlet->modify (line 144 of verlet.cpp) and
one came out whilst the other got stuck in xcm.
thanks for the effort. i think at this moment, we first need to have an
input deck that works properly on a single processor and this doesn't look
like it. if there are all kinds of bad things happening to atoms, then it
is difficult to tell, whether there is a hiccup in LAMMPS or just bad
system dynamics.
axel.