[gmx-users] Problems with installing GROMACS with MPI

[Дмитрий Дударев]

Dear all,

I am trying to install GROMACS on the server with CUDA and MPI support. I
figured it out with CUDA (tried it in the MPI-less version). The problem
was with the memory that 'gmx mdrun' uses out of GPUs - 'nvidia-msi' showed
that only ~350 MB (of ~16000MB) has been used by 'mdrun' from 2
accelerators, which is... low result I guess?

Nevertheless, I tried to install GROMACS with MPI support now, mayve this
will improve the final result. As a result - many tests during 'make check'
procedure failed and after installation when I try 'gmx_mpi mdrun -h' it
shows options for the command, but with such message in the end:

simulacrum-PowerEdge-R740:03726] *** Process received signal ***
[simulacrum-PowerEdge-R740:03726] Signal: Segmentation fault (11)
[simulacrum-PowerEdge-R740:03726] Signal code: Address not mapped (1)
[simulacrum-PowerEdge-R740:03726] Failing at address: 0x99
[simulacrum-PowerEdge-R740:03726] [ 0]
/lib/x86_64-linux-gnu/libpthread.so.0(+0x12890)[0x7f9a33b44890]
[simulacrum-PowerEdge-R740:03726] [ 1]
/usr/lib/x86_64-linux-gnu/libopen-pal.so.20(opal_hwloc_base_free_topology+0x76)[0x7f9a33279ca6]
[simulacrum-PowerEdge-R740:03726] [ 2]
/usr/lib/x86_64-linux-gnu/libopen-pal.so.20(+0x6f830)[0x7f9a33276830]
[simulacrum-PowerEdge-R740:03726] [ 3]
/usr/lib/x86_64-linux-gnu/libopen-pal.so.20(mca_base_framework_close+0x71)[0x7f9a332551c1]
[simulacrum-PowerEdge-R740:03726] [ 4]
/usr/lib/x86_64-linux-gnu/libopen-pal.so.20(opal_finalize+0x95)[0x7f9a3322e365]
[simulacrum-PowerEdge-R740:03726] [ 5]
/usr/lib/x86_64-linux-gnu/libmpi.so.20(ompi_mpi_finalize+0x69d)[0x7f9a3496638d]
[simulacrum-PowerEdge-R740:03726] [ 6] gmx_mpi(+0x93de6)[0x562531385de6]
[simulacrum-PowerEdge-R740:03726] [ 7] gmx_mpi(+0x8be33)[0x56253137de33]
[simulacrum-PowerEdge-R740:03726] [ 8]
/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xe7)[0x7f9a33762b97]
[simulacrum-PowerEdge-R740:03726] [ 9] gmx_mpi(+0x8f88a)[0x56253138188a]
[simulacrum-PowerEdge-R740:03726] *** End of error message ***
Segmentation fault (core dumped)

How can I solve this issue? I'm not very expirienced user, that's why I'm
writing here. Hope for the solution soon.

-- 
With best wishes,
Dudarie Dmytro