[QE-users] crash of fs.x

[patrizio.graziosi at cnr.it]

  Dear all,

  since I have noticed the attachments might have been removed, I  
repeat the message and report the text files at the end of the message  
instead of as attachments. So sorry for the duplicate.

  I am experiencing crashes of fs.x when using versions 6.3, 6.4, 6.5  
and 6.6 either.
  I'm running Quantum Espresso on the supercomputing cluster "galileo"  
at CINECA, Italy, equipped with Intel Xeon E5-2697 v4 (Broadwell)  
nodes, 1022 36-core compute nodes, each one contains 2 18-cores Intel  
Xeon E5-2697 v4 (Broadwell) at 2.30 GHz.


  I do as follows, all the input files and job submission scripts are  
reported at the end of themessage:


  1) run a scf calculation

  2) run a nscf calculation

  3) run fs.x


  at step 3) QE crashes and I get in the .out file

      Error in routine fill_fs_grid (1):
      cannot locate  k point


  and in the CRASH file

      task #        19
      from fill_fs_grid : error #         1
      cannot locate  k point



  Please, note that I asked to colleagues of mine at the University of  
Warwick, where they use QE on a HPC infrastructure (Tinis) with Lenovo  
NeXtScale nx360 M5 servers with 2 x Intel Xeon E5-2630 v3 2.4 GHz  
(Haswell) 8-core processors, to run QE with the very same input files,  
and they
  told me that everything run smoothly with NO crash...


  So, I suppose I'm doing a silly error in the submission script, or,  
as suggested by my colleagues at Warwick, there is a compiling issue  
on the CINECA machine I'm using.


  Can you support me in understanding where or what I have to look at,  
in order to deal with this issue, please?



  BTW in step 2) I also used occupations = 'tetrahedra' or 'fixed',  
with the same crash for fs.x


  I thank you in advance for your help ,


  Patrizio



1) scf calculation

  input file

  &control
     prefix='Mg3Sb2',
     pseudo_dir = './',
     outdir='./'
     wf_collect=.true.
     etot_conv_thr = 1.0d-8,
/
  &system
     ibrav=  4,
     celldm(1) = 8.6767152,
     celldm(3) = 1.5818114,
     nat=  5,
     ntyp= 2,
     ecutwfc = 100.0,
  /
  &electrons
     conv_thr = 1.0d-8,
     mixing_beta = 0.7,
  /
ATOMIC_SPECIES
  Mg  24.305  Mg.upf
  Sb  121.76  Sb.upf
ATOMIC_POSITIONS crystal
  Mg 0      0      0
  Mg 1/3 2/3 0.3679670359
  Mg 2/3 1/3 0.6320329641
  Sb 1/3 2/3 0.7745210172
  Sb 2/3 1/3 0.2254789828
K_POINTS automatic
    13 13 9 0 0 0



  submission script

#!/bin/bash
#SBATCH --time=24:00:00        # Walltime in hh:mm:ss
#SBATCH --nodes=4              # Number of nodes
#SBATCH --ntasks-per-node=6   # Number of MPI ranks per node
#SBATCH --cpus-per-task=6      # Number of OpenMP threads for each MPI  
process/rank
#SBATCH --mem=118000           # Per nodes memory request (MB)
#SBATCH --account=IscrC_MEETH
#SBATCH --job-name=mg3sb2
#SBATCH --partition=gll_usr_prod

module purge
module load profile/phys
module load autoload qe/6.6

export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK
export MKL_NUM_THREADS=${OMP_NUM_THREADS}

mpirun pw.x -npool 24 -input scf.in > Mg3Sb2.out



2) nscf calculation

  input file

  &control
     prefix='Mg3Sb2',
     pseudo_dir = './',
     outdir='./'
     wf_collect=.true.
     etot_conv_thr = 1.0d-8,
     calculation = 'nscf',
/
  &system
     ibrav=  4,
     celldm(1) = 8.6767152,
     celldm(3) = 1.5818114,
     nat=  5,
     ntyp= 2,
     ecutwfc = 100.0,
  /
  &electrons
     conv_thr = 1.0d-8,
     mixing_beta = 0.7,
  /
ATOMIC_SPECIES
  Mg  24.305  Mg.upf
  Sb  121.76  Sb.upf
ATOMIC_POSITIONS crystal
  Mg 0      0      0
  Mg 1/3 2/3 0.3679670359
  Mg 2/3 1/3 0.6320329641
  Sb 1/3 2/3 0.7745210172
  Sb 2/3 1/3 0.2254789828
K_POINTS automatic
    61 61 41 0 0 0


  submission script, as above but last line

mpirun pw.x -npool 24 -input nscf.in > Mg3Sb2_nscf.out


3) FS calculation

  input file

&fermi
	outdir = './',
	prefix = 'Mg3Sb2',
	DeltaE = 3
/


  submission script, as above but last line

mpirun fs.x -npool 24 -input fs.in > Mg3Sb2_fermi.out