From: McGuire, Kelly (mcg05004_at_byui.edu)
Date: Thu Mar 21 2019 - 12:25:34 CDT
Hey Brian, no worries. I have a machine file written to the orca directory. This machinefile has the available nodes that orca can use. The file should have the same name as you input file, but end in .nodes
We use a file called generate_pbs_nodefile to generate this list. You might have one of these.
In your orca input file, you should have:
%pal
nprocs 16
end
We use SLURM. My submit script is shown below.
1) I specify the path to our openmpi (ORCA is stubborn on the openmpi version, make sure you have the one that goes with your version of ORCA)
2) The #Machinefile section generates the nodelist and prints the list in a .nodes file that is written in our orca directory and my directory where my input and other files are kept
3) I load orca and openmpi with the module load section
4) The full path to our ORCA is given under #Submit. A copy of your input should be placed in the orca directory, but I run my submit file from my personal directory which has the input file there as well
#!/bin/bash
#SBATCH -C rhel7
#SBATCH --time=1-00:00:00 # walltime
#SBATCH -N 2 --ntasks-per-node=16
#SBATCH --mem=64G # memory per CPU core
##SBATCH -w, --nodelist=m8-17-6
# VARIABLES
# export OMP_NUM_THREADS=$SLURM_CPUS_ON_NODE
# OPENMPI Paths
export PATH=path_to_openmpi/bin:$PATH
export LD_LIBRARY_PATH=path_to_openmpi/lib:$LD_LIBRARY_PATH
# Machinefile for ORCA
hostname -s
cat `path_to_gen_pbs_file/generate_pbs_nodefile`
egrep -v "^`hostname -s`$" `path_to_gen_pbs_file/generate_pbs_nodefile` > path_to_orca/CuAMTIDA.nodes
egrep -v "^`hostname -s`$" `path_to_gen_pbs_file/generate_pbs_nodefile` > path_to_personal_directory/CuAMTIDA.nodes
module load path_to_orca/orca
module load path_to_openmpi/bin/
# CUDA TOOLKIT PATH
export CUDA_HOME=/zapps7/cuda/8.0.61/bin/
# SUBMIT
full_path_to_orca/orca CuAMTIDA.inp > CuAMTIDA.out
Kelly L. McGuire
PhD Candidate
Biophysics
Department of Physiology and Developmental Biology
Brigham Young University
LSB 3050
Provo, UT 84602
________________________________
From: Bennion, Brian <bennion1_at_llnl.gov>
Sent: Thursday, March 21, 2019 11:07 AM
To: namd-l_at_ks.uiuc.edu; McGuire, Kelly
Subject: RE: QMMM Frequency Calculation
Kelly,
Sorry this is off list, but I was curious if I missed how you got orca parallelized over the 16 cores?
Brian
From: owner-namd-l_at_ks.uiuc.edu <owner-namd-l_at_ks.uiuc.edu> On Behalf Of McGuire, Kelly
Sent: Thursday, March 21, 2019 9:55 AM
To: namd-l_at_ks.uiuc.edu
Subject: namd-l: QMMM Frequency Calculation
With ORCA working in parallel on 16 processors now, my QMMM is going fast. But, if I add in frequency calculations, the QMMM simulations slows down, predictably. I don't necessarily need the thermochemistry for every step any way. So, my thought is watch the trajectory, and any steps where there is something "interesting" happening, such as a covalent bond forming, save the frame before the bond formation as a PDB and the frame where the bond has formed as a PDB, and then run a frequency calculation on those frames/structures, I guess holding the MM region frozen, and that would give me thermochemistry data for the important parts of my QMMM. What are your thoughts on this?
Kelly L. McGuire
PhD Candidate
Biophysics
Department of Physiology and Developmental Biology
Brigham Young University
LSB 3050
Provo, UT 84602
This archive was generated by hypermail 2.1.6 : Tue Dec 31 2019 - 23:20:38 CST