[QE-users] Possible resolution of previous post on mysterious ph.x error

Wed Oct 26 22:34:46 CEST 2022

Dear Quantum Espresso forum,
      Yesterday I noted a mysterious error in ph.x (Version 7.1) observed
for NaCl using the primitive cell.    The error was quite strange, usually
mentioning an error in  cdiaghg which is a routine in the
LAXlib directory.   I now think that it has to do with requesting too many
cores for my calculation because my error can be removed by requesting
fewer cores.     As noted in the original post,   the error did not appear
in the QE versus 6.3, so presumably LAXlib files have been changed in the
7.1 version.    I very much appreciate the fact that QE seems to adapt to
the number of nodes and cpus per node available in each run, but perhaps I
need to increase my understanding of how this works for the various
programs.   There was no problem running pw.x or ph.x for the gamma point
with 32 cores,   but ph.x failed for q /=0.    Both pw.x and ph.x  with
all q vectors seem to have run correctly with 8 cores.     Are there some
guidelines for choosing how many nodes/cores to request  ?   In any case,
  thanks for listening.       Sincerely,    Natalie Holzwarth

N. A. W. Holzwarth                                       email:
natalie at wfu.edu
Department of Physics                                  web:
http://www.wfu.edu/~natalie
Wake Forest University                                 phone:
1-336-758-5510
Winston-Salem, NC 27109 USA                     office: Rm. 300 Olin
Physical Lab

---------- Forwarded message ---------
From: Holzwarth, Natalie <natalie at wfu.edu>
Date: Tue, Oct 25, 2022 at 4:50 PM
Subject: Very strange issue with ph.x in QE 7.1 which appeared OK in QE 6.3
To: Quantum Espresso users Forum <users at lists.quantum-espresso.org>

Dear Quantum Espresso Forum,

I have been using ph.x from QE 7.1 for quite a few low-symmetry materials
but for several simple fcc crystals, the program
fails to diagonalize the first non-gamma point phonon matrix (see 7.1G.out)
file.   However I ran the same program with QE 6.3
which diagonalizes all of the phonon modes of NaCl.  In this test,   I am
using pseudos from SSSP.    The input file is pasted below
and the two output files from ph.x are hopefully attached.     Has anyone
else seen this issue?.
Thanks very much for any advice about this, Natalie

input file for vc-relax and phonon calculation:
------------------------------------------------------------
cat runQE.slurm
#!/bin/tcsh
#
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=32
#SBATCH --cpus-per-task=1
#SBATCH --account="natalieGrp"
#SBATCH --output="JOB-%j.o"
#SBATCH --error="JOB-%j.e"
#SBATCH --mail-user=natalie at wfu.edu
#SBATCH --mail-type=BEGIN,END,FAIL
#SBATCH --time=0-300:00:00
#SBATCH --mem=96gb
#SBATCH --partition=large
umask 002
# Note: SLURM has no batch input for cputime, excluding.
#
#
echo 'hostname' `/bin/hostname`
echo 'job directory' `pwd`
#
setenv TMPDIR /scratch/$SLURM_JOBID
echo 'Reset TMPDIR for this job to ' $TMPDIR

module load apps/quantum-espresso/7.1
set PW=pw.x
set PH=ph.x

#NOTE:SLURM defaults to running jobs in the directory where submitted;
#NOTE:Consider --workdir directive instead; and check functionality!
cd ${SLURM_SUBMIT_DIR}

cat > PSI.in << EOF
&CONTROL
  calculation = "vc-relax",
  pseudo_dir  =
'/deac/phy/natalieGrp/natalie/wfurc9/EL7/QE/Downloadedpseudos'
  verbosity   = "high",
  outdir      = "$TMPDIR/",
  prefix      = 'PSI',
  restart_mode = 'from_scratch',
  nstep = 300,
  dt = 20,
  forc_conv_thr = 1.0D-5,
  etot_conv_thr = 1.0D-6,
  tstress = .true.,
  tprnfor = .true.,
/
&SYSTEM
ibrav       =  2,
celldm(1)   = 10.6,
nat         = 2,
ntyp        = 2,
nosym       =.FALSE.,
use_all_frac = .TRUE.,
ecutwfc     = 81.d0,
/
&ELECTRONS
  conv_thr    = 1.D-7,
  electron_maxstep = 200,
/
&IONS
/
&CELL
  cell_dynamics='bfgs',
  wmass = 1.00,
  press = 0.0,
/
ATOMIC_SPECIES
 Na  22.989769   Na_ONCV_PBEsol-1.0.upf
 Cl  35.45   Cl.pbesol-n-rrkjus_psl.1.0.0.UPF

ATOMIC_POSITIONS {crystal}
 Na      0.0   0.0    0.0
 Cl      0.5   0.50    0.50

K_POINTS AUTOMATIC
 8 8 8 0 0 0
EOF

mpirun $PW  -in   PSI.in  > PSI.out

cat > PSI.phG.in << EOF
phonons

&inputph
outdir = '$TMPDIR',
prefix = 'PSI',
epsil = .true.,
ldisp = .true.,
fildyn = 'dyn.G',
tr2_ph = 1.0d-14,
start_q=1,
last_q=8,
nq1 = 4,
nq2 =4,
nq3 = 4,
/

EOF
mpirun $PH -in PSI.phG.in > PSI.phG.out

ls -Flag $TMPDIR

'rm' -r $TMPDIR

/usr/local/bin/slurm_mem_report -v
------------------------------------------------------------
N. A. W. Holzwarth                                       email:
natalie at wfu.edu
Department of Physics                                  web:
http://www.wfu.edu/~natalie
Wake Forest University                                 phone:
1-336-758-5510
Winston-Salem, NC 27109 USA                     office: Rm. 300 Olin
Physical Lab
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20221026/699e434e/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 6.3phG.out
Type: application/octet-stream
Size: 20322 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20221026/699e434e/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 7.1phG.out
Type: application/octet-stream
Size: 22899 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20221026/699e434e/attachment-0001.obj>