I am catching a "MPI process rank 0 (n0, p25243) caught a SIGSEGV" error at the end of my lammps cuda run. Am I going to have to debug the lmp_openmpi executable to find out what's wrong? The run appears to be okay, apparently it is not exiting correctly?
- Randall
I am catching a "MPI process rank 0 (n0, p25243) caught a SIGSEGV" error
at the end of my lammps cuda run. Am I going to have to debug the
lmp_openmpi executable to find out what's wrong? The run appears to be
okay, apparently it is not exiting correctly?
yup. you can try to find out where it is failing by
compiling your executable with debug info enabled
and then using ulimit -c unlimited to allow core dumps
and then inspect the core dump(s) using gdb.
i've described the process in the past in detail,
and thus you should be able to find the instructions
in the mailing list archives.
cheers,
axel.