Memory Error
Just Got Here
3:33:55 AM PDT - Thu, Aug 1st 2019
Dear all,

I'm trying to run a geometry optimization of a quite big system and I'm getting the error movecs_write: ma failed 9955. I read that this problem is due to the Memory I'm asking for the calculation.
Is the first time I'm trying to run a nwchem calculation and I don’t' know how the memory specifications should be.

The computer I'm using has the following specifications:
48 Cores x Node and 1.880 Gb x Core.
I'm running the calculation using an entire Node.

Is there any rational way to consider the memory specifications?

Pleas find attached below the Input I'm using.

Thank you.

Best regards,
Gerard

title "MOF"

memory stack 469 mb heap 100 mb global  1306 mb



#scratch_dir /scratch



geometry units angstroms
Cr 8.547900000000 -6.708900000000 45.405000000000
O 6.960500000000 -6.802600000000 46.260600000000
... "up to 1166 atoms"
end
charge 0

basis
  1. BASIS SET: (3s) -> [2s]
H S
     0.5447178000E+01       0.1562849787E+00

     0.8245472400E+00       0.9046908767E+00
H S
     0.1831915800E+00       1.0000000
  1. BASIS SET: (6s,3p) -> [3s,2p]
C S
     0.1722560000E+03       0.6176690738E-01

     0.2591090000E+02       0.3587940429E+00

     0.5533350000E+01       0.7007130837E+00
C SP
     0.3664980000E+01      -0.3958951621E+00       0.2364599466E+00

     0.7705450000E+00       0.1215834356E+01       0.8606188057E+00
C SP
     0.1958570000E+00       0.1000000000E+01       0.1000000000E+01
  1. BASIS SET: (6s,3p) -> [3s,2p]
O S
     0.3220370000E+03       0.5923939339E-01

     0.4843080000E+02       0.3514999608E+00

     0.1042060000E+02       0.7076579210E+00
O SP
     0.7402940000E+01      -0.4044535832E+00       0.2445861070E+00

     0.1576200000E+01       0.1221561761E+01       0.8539553735E+00
O SP
     0.3736840000E+00       0.1000000000E+01       0.1000000000E+01
  1. BASIS SET: (12s,9p,3d) -> [5s,4p,2d]
Cr S
     0.2798294400E+04       0.6382379506E-01

     0.4231369600E+03       0.3770839708E+00

     0.9243886100E+02       0.6809888473E+00
Cr SP
     0.1202805620E+03      -0.1177790010E+00       0.1398781830E+00

     0.2603727080E+02       0.1014311230E+00       0.5559834482E+00

     0.7844172480E+01       0.9571981377E+00       0.4748183482E+00
Cr SP
     0.7793275650E+01      -0.2888567131E+00       0.2218470920E-01

     0.2497196130E+01       0.3351146841E+00       0.4616249951E+00

     0.8051418920E+00       0.8502480973E+00       0.6145386131E+00
Cr SP
     0.7039205640E+00      -0.2322507691E+00       0.1799644840E-03

     0.8616195330E-01       0.1093671325E+01       0.9999447611E+00
Cr SP
     0.3219881930E-01       0.1000000000E+01       0.1000000000E+01
Cr D
     0.9625338600E+01       0.2655959100E+00

     0.2362264300E+01       0.8521556801E+00
Cr D
     0.5770944000E+00       1.0000000
END
driver

  maxiter 1000

end

dft

  mult 1

  iterations 256

  xc becke88 perdew86

  grid fine

  direct

end

task dft optimize

Forum Vet
9:52:42 AM PDT - Thu, Aug 1st 2019
number of basis functions
Gerard,
Could you please post the number of basis functions used?
You can find in the output file by searching for the line
AO basis - number of functions:

This number will give me an estimate of the memory requirements.

By the way, are you using just a single process in your run?

Thanks

Just Got Here
2:32:58 AM PDT - Fri, Aug 2nd 2019
Yes, no problem,
AO basis - number of functions: 9955
                    number of shells:  5339
Yes, I'm using a single process.
Please find attached here the luncher:

  1. !/bin/bash
  2. SBATCH --ntasks=48
  3. SBATCH --time=12:00:00
  4. SBATCH --cpus-per-task=1
  5. SBATCH --output=slurm%J.out
  6. SBATCH --error=slurm%J.err
module load nwchem
  1. module load fabric
time mpirun nwchem input.nw

Thanks.

Forum Vet
9:15:53 AM PDT - Fri, Aug 2nd 2019
Sorry to contradict you, but the line

SBATCH --ntasks=48

would indicate that mpirun should spawn 48 processes. (Anyhow, you can check the actual number of processes by searching for the nproc string in the NWChem output)

Since your input results in 9955 basis functions, you need at least 756 mbytes of stack memory (9955*9955*8/1024/1024=756).
I would set it to 1400 mb to be on the safe side. If the program crashes early on with a memory error, you might try to reduce the number of tasks from 48 to 24



memory stack 1400 mb heap 100 mb global  1306 mb


Forum >> NWChem's corner >> Running NWChem