The computational cluster I have been submitting jobs is switching clusters from one that supports 24 ncpus to 36 ncpus. The problem is, now when I run my jobs it produces the following error:
36 OpenMP threads were requested. Since the non-bonded force buffer reduction
is prohibitively slow with more than 32 threads, we do not allow this. Use 32
or less OpenMP threads.
I have asked the computational cluster individuals for help, but they sent me the documentation to GROMACS:
https://manual.gromacs.org/documentation/2018/user-guide/environment-variables.html
Here is my relevant code:
-------------------------------------------------------------------------------------------------------------------
#!/bin/bash
#PBS -l select=1:ncpus=36:mpiprocs=12:mem=1GB,walltime=01:00:00
(NOTE: This value MUST be 36. If I try anything else, the job is immediately rejected)
#PBS -q qprod
#PBS -N test
HOME_DIR=/scratch/.../…
cd $HOME_DIR
ml GROMACS/2016.5-intel-2017c-hybrid-single-PLUMED-2.3.8
gmx_mpi mdrun -ntomp 36 -s energymin.tpr -deffnm energymin -v &> energymin.mdrun
--------------------------------------------------------------------------------------------------------------------
If anyone has any suggestions, I would be extremely happy to hear them! I am slightly out of my depth on this one…
Thank you so much ahead of time,
Michael