mpirun nwchem on Windows Subsystem for Linux?


Clicked A Few Times
I have nwchem 6.8 working properly in serial mode on Ubuntu 18.10 via Windows Subsystem for Linux. I installed nwchem via "sudo apt install nwchem". I use a simple input file water.nw for geometry optimization for water. The following commands all work:

nwchem water.nw > water.out
mpirun -np 1 nwchem water.nw > water_MPI.out
mpirun -np 1 /usr/bin/nwchem water.nw > water_MPI.out


But changing the number of processors from 1 to 2 causes the calculation to stall. In water_MPI.out, there is only one line, "argument 1 = water.nw". In the terminal, the following error shows up, and then the calculation hangs without ever completing.

--------------------------------------------------------------------------
WARNING: Linux kernel CMA support was requested via the
btl_vader_single_copy_mechanism MCA variable, but CMA support is
not available due to restrictive ptrace settings.

The vader shared memory BTL will fall back on another single-copy
mechanism if one is available. This may result in lower performance.

  Local host: DESKTOP-89MM0K0
--------------------------------------------------------------------------
[DESKTOP-89MM0K0:16805] 1 more process has sent help message help-btl-vader.txt / cma-permission-denied
[DESKTOP-89MM0K0:16805] Set MCA parameter "orte_base_help_aggregate" to 0 to see all help / error messages


Does anyone have any ideas what is causing this?
mpirun --mca orte_base_help_aggregate 0 -np 2 /usr/binnwchem water.nw > water_MPI.out
and
mpirun -v -np 2 /usr/binnwchem water.nw > water_MPI.out
don't give any more error messages to help debug what is going on.

Forum Vet

echo 0 | sudo tee /proc/sys/kernel/yama/ptrace_scope


https://github.com/Microsoft/WSL/issues/3397#issuecomment-417876710

https://groups.io/g/OpenHPC-users/topic/openmpi_and_shared_memory/16489081

Clicked A Few Times
Hi Edoapra,

Thanks for the helpful information! I had come across similar links when googling my error message, and have set the ptrace_scope file to 0. Now, "mpirun -np 2 /usr/bin/nwchem water.nw > water_MPI.out" results in no errors showing up in the terminal, but water_MPI.out still only contains "argument 1 = water.nw", and the calculation appears to be stuck for at least an hour, whereas serial execution finishes in less than a second.

-Sibo


Forum >> NWChem's corner >> Running NWChem