Problem with input for NWChem 6.1.1 and 6.5


Clicked A Few Times
Hello,
I successfully used NWChem 6.1.1 on CRAY using this input:
Quote:INPUT
START Al2O3-newcluster-OD_BS
memory stack 1024 mb heap 1024 mb global 2048 mb noverify
geometry units angstroms noautosym noautoz
//
// list of 303 atoms
//
END
BASIS "ao basis" PRINT
  1. BASIS SET: (3s) -> [2s]
H S
     5.4471780              0.1562850        
0.8245470 0.9046910
H S
     0.1831920              1.0000000        
  1. BASIS SET: (6s,3p) -> [3s,2p]
O S
   322.0370000              0.0592394        
48.4308000 0.3515000
10.4206000 0.7076580
O SP
     7.4029400             -0.4044530              0.2445860        
1.5762000 1.2215600 0.8539550
O SP
     0.3736840              1.0000000              1.0000000        
  1. BASIS SET: (9s,6p) -> [4s,3p]
Al S
   775.7370000              0.0668347        
116.9520000 0.3890610
25.3326000 0.6694680
Al SP
    29.4796000             -0.1079020              0.1175740        
6.6331400 0.1462450 0.4611740
1.7267500 0.9237300 0.6055350
Al SP
     0.9461600             -0.3203270              0.0519383        
0.2025060 1.1841200 0.9726600
Al SP
     0.0639088              1.0000000              1.0000000        
END

set fock:replicated logical .false.
set geometry:actlist 1:207
DFT
print "convergence"
direct
XC b3pw91
iterations 9999
convergence energy 1e-6
convergence damp 30
END
DRIVER
MAXITER 999
xyz Al2O3_newcluster
EPREC 1e-6
END
task dft optimize
title "Optimized geometry"
dft
XC b3pw91
mulliken
iterations 9999
end
task dft energy

In NWChem 6.1.1, in result i get number of created .xyz files and calculated Total DFT energy. OUTPUT
Using NWChem 6.5 program creates only number of .gridpts files and doesn't calculate geometry. I use same or more processors for both programs, while NWChem 6.5 is running on more powerfull Xeons (16 processors per node). OUTPUT
Here is submission script for Xeon:
Quote:username
#!/bin/bash
 #SBATCH -J water_test                # jobname
#SBATCH -A CORRADEF # project name
#SBATCH -N 32 # number of nodes
#SBATCH -n 16 # number of tasks
#SBATCH -o %j.out # strout filename (%j is jobid)
#SBATCH -e %j.err # stderr filename (%j is jobid)
#SBATCH -t 24:00:00 # execute time
# mpirun -np
export ARMCI_DEFAULT_SHMMAX=8192
export NWC_RANKS_PER_DEVICE=0
export OMP_NUM_THREADS=4
export MIC_USE_2MB_BUFFER=16K
export ARMCI_OPENIB_DEVICE=mlx4_0
module load intel bullxmpi srun

mpirun -np 16 /csc/scratch2/kotomin/nwchem nwchem.nw


I just noticed that NWChem6.5 output shows that I use only 16 processors, but I see in system that process is running on 32 nodes.
So where the problem might be?

thank you!

Forum Vet
The line
mpirun -np 16 /csc/scratch2/kotomin/nwchem nwchem.nw
is causing 16 processes to be used.

Clicked A Few Times
I typed 512 in that line, launched NWChem on 32 nodes with 16 processors on each and got this: ERROR

Forum Vet
Could you please replace the "mpirun ... nwchem" line with the following line and send me the complete output

mpirun /bin/hostname

Clicked A Few Times
Here is the result: HOSTNAME

Submission script was
Quote:username
#!/bin/bash

#SBATCH -J water_test                # jobname
#SBATCH -A CORRADEF # project name
#SBATCH -N 16 # number of nodes
#SBATCH -n 16 # number of tasks
#SBATCH -o %j.out # strout filename (%j is jobid)
#SBATCH -e %j.err # stderr filename (%j is jobid)
#SBATCH -t 24:00:00 # execute time
# mpirun -np
export ARMCI_DEFAULT_SHMMAX=8192
export NWC_RANKS_PER_DEVICE=0
export OMP_NUM_THREADS=4
export MIC_USE_2MB_BUFFER=16K
export ARMCI_OPENIB_DEVICE=mlx4_0
module load intel bullxmpi srun

mpirun /bin/hostname

Forum Vet
Could you upload the complete error and output files?

Clicked A Few Times
Unfortunately, i don't have separate error files.
Here is complete output for launching NWChem with mpirun -np 512 .... OUTPUT
Complete output for mpirun /bin/hostname was in my previous post.


Forum >> NWChem's corner >> Running NWChem