How to run NAMD 2.14 on cores from multiple nodes in slurm? need batch file please.
this is the one i am using:
#!/bin/sh
#SBATCH --job-name=grp_prod
#SBATCH --partition=cpu
#SBATCH --ntasks=48
#SBATCH --cpus-per-task=1
#SBATCH --time=30-50:20:00
mpirun -np 48 /lfs01/workdirs/cairo029u1/namd2_14/NAMD_2.14_Linux-x86_64-multicore/namd2 +auto-provision step4_equilibration.inp > eq.log
this batch file produces this attached log file (the first part only is attached)
Sanjay Kharche it does not work properly. the log file produced has repeated sections in it. so i think each core is running on its own or something like that
The log file appears to indicate a shared memory parallelization, not MPI. I am assuming you have compiled using the appropriate shared memory compiler (e.g. openmp) and have provided the correct flags at run time. Also, the for loops in the code need to have a SMP call.
Sanjay Kharche i am just a user of NAMD so i cant understand most of your last answer. could you explain it more, please?
the command i used in running is
mpirun -np 48 /lfs01/workdirs/cairo029u1/namd2_14/NAMD_2.14_Linux-x86_64-multicore/namd2 +auto-provision step4_equilibration.inp > eq.log
the commands i used in the batch file are written in the question
There are two ways to parallelize your program: MPI and OpenMP. The simulation is executed using two different wrappers. MPI is executed using mpiexec. I suspect from your log file that your program is OpenMP compiled.
Having said all this, v2.14 appears to be MPI enabled - can you somehow double check if the source was compiled using MPI?
You can try this job.sh :
#!/bin/bash #SBATCH --job-name=myNAMDbinary #SBATCH --output=myLog.log #SBATCH --error=myLog.err #SBATCH --nodes=1 # your log file suggests you have 1 node, skip this line if need #SBATCH --ntasks-per-node=22 # write however many threads you have.
# load modules you need here.
export OMP_NUM_THREADS=22 /lfs01/workdirs/cairo029u1/namd2_14/NAMD_2.14_Linux-x86_64-multicore/namd2 +auto-provision step4_equilibration.inp > eq.log
the v2.14 of namd is the one downloaded from their website. this is its name on the website :
Linux-x86_64-multicore (64-bit Intel/AMD single node)
does this mean that it works on a single node?
these are all the linux versions on their website for v2.14:
- Linux-x86_64-multicore (64-bit Intel/AMD single node)
- Linux-x86_64-multicore-CUDA (NVIDIA CUDA acceleration)
- Linux-x86_64-netlrts (Multi-copy algorithms, single host or ethernet)
- Linux-x86_64-netlrts-smp-CUDA (Multi-copy algorithms, single process per copy)
- Linux-x86_64-verbs (InfiniBand, no MPI needed, supports multi-copy algorithms)
- Linux-x86_64-verbs-smp (InfiniBand, no MPI needed, supports multi-copy algorithms)
- Linux-x86_64-verbs-smp-CUDA (InfiniBand, no MPI needed, supports multi-copy algorithms)
- Linux-KNL-multicore (Intel Xeon Phi KNL processor single node)
I have no way of telling from binary name alone.
A node is a computer that has many cores (processors) connected in some way, and connected to memory in some way.
My choice would be to obtain the source code and compile yourself. By doing so, you will ensure that all the parallel libraries (MPI and openmp) are visible to your binary (compiled program).
Good luck.
- Badges
- Science topic
- Similar topics
- Filing