Asked to delete not existent file: <filename>.abo_PRT1

Hello,

I keep getting this annoying message during RF run (abinit 9.2.2). Sometimes the code gets past this part and others the reserved time on my cluster runs out for some reason?

Asked to delete not existent file: <filename>.abo_PRT1
Asked to delete not existent file: <filename>.abo_PRT2
Asked to delete not existent file: <filename>.abo_PRT3
and so on...

I was doing RF calculations with 4x4x4 k-grid (SOC included: nspinor=2 nspinden=1 nsppol=1), this wasn’t an issue, but when I try with a 8x8x8 k-grid that message either stops or prolongs the calculation.

I couldn’t find this related topic anywhere else, so I’m sorry in advance if there already is a solution to this issue.

Kind regards,
Milan Jocic

Hi mjocic,

we need a bit more information, on your input file and what you want to do in the perturbations. FYI if you set nspinor 1 the SOC can not be applied.

The files it complains about are log files for multiple perturbation parallelization. This may also not be what you intended to do: it’s more efficient to run 1 perturbation at a time and parallelize over k bands etc.

Have a look at the abipy tutorials as well (it does all this automatically). At any rate this is not the actual problem, but may be a clue.

Hello,

Sorry, I mistyped for SOC, I use nspinor=2.

I use one q-point per process:
These are the options I use to run the code when q!=[0 0 0]:

# PARALLEL OPTIONS
#   paral_rf   1
#  paral_kgb   1
   timopt        -3
# NUMBER OF DATASETS
   ndtset   3
# Set 1: ground state self-consistency
   getwfk1    0            # Cancel default
   kptopt1    1            # Auto gen using symmetry (1-all 2-just time 3-NONE)
     nqpt1    0            # Cancel default
   tolvrs1    ${TOLVRSGS}      # SCF stopping criterion (modify default)
   rfphon1    0            # Cancel default
  
  paral_kgb1  1            # PARALLEL OPTION
  autoparal1  1            # PARALLEL OPTION

  paral_rf3   1            # PARALLEL OPTION
  autoparal3  1            # PARALLEL OPTION
# Set 2: non self-consistent calculation
   getden2     1
     iscf2    -2
     nqpt2     1 
   nbdbuf2     4
   tolwfr2    ${TOLWFR}
     kptopt2     3
   
   paral_kgb2  1            # PARALLEL OPTION
   autoparal2  1            # PARALLEL OPTION


## COMMON VARIABLES
   getwfk   1                  #  Use GS wave functions from dataset1
     nqpt   1                  #  number of q points=1 per job/nod
      qpt   ${q_point}         
                               #  qpoint from file QLIST
   kptopt   3                  #  use no symmetries
   tolvrs   ${TOLVRS}             #  tolerance for Vrs
   rfphon   1                  #  resp. func. phonon calc
  rfatpol   1 5                #  all atoms displacement
    rfdir   1 1 1              #  consider all directions

and when q==[0 0 0]

# PARALLEL OPTIONS
#  paral_rf    1
# paral_kgb    1
    timopt   -3
# NUMBER OF DATASETS
    ndtset   3 
# Set 1: ground state self-consistency
#   prtden1    1
   getwfk1    0            # Cancel default
   kptopt1    1            # Auto gen using symmetry (1-all 2-just time 3-NONE)

     nqpt1    0            # 
   tolvrs1    ${TOLVRSGS}      # SCF stopping criterion (modify default)
   rfphon1    0            # Cancel default

  paral_kgb1  1            # PARALLEL OPTION
  autoparal1  1            # PARALLEL OPTION
# Set 2: Response function calculation of d/dk wave function
     iscf2   -3            # Need this nscf option for ddk
   kptopt2    2            # Use just time-rev symmetry
   rfphon2    0            # Cancel delault
   rfelfd2    2            # Calculate ddk only
   tolwfr2    ${TOLWFR}    # Use wf residual criterion instead
   nbdbuf2    4            # Exclude last few bands from tolwrf test

   paral_rf2  1            # PARALLEL OPTION
   autoparal2 1            # PARALLEL OPTION
# Set 3: Response function calculation of Q=0 phonons and el.field
   getddk3    2            # ddk wave functions from dataset2
   kptopt3    2            # 
   rfelfd3    3            # Electric-field perturbation response only
   tolwfr3    ${TOLWFR}
   nbdbuf3    2
   paral_rf3  1            # PARALLEL OPTION
   autoparal3 1            # PARALLEL OPTION
# COMMON VARIABLES
   getwfk   1                  #  Use GS wave functions from dataset1
     nqpt   1                  #  number of q points=1 per job/nod
      qpt   ${q_point}         
                               #  qpoint from file QLIST
   kptopt   3                  #  use no symmetries
   tolvrs   ${TOLVRS}             #  tolerance for Vrs
   rfphon   1                  #  resp. func. phonon calc
  rfatpol   1 5                #  all atoms displacement
    rfdir   1 1 1              #  consider all directions

Hello again,

The _PRT file is made to parallelize over perturbations as we mentioned before, to make a separate output and log file for each perturbation. I don’t know why it hangs there, could be an issue with disk IO or some conflict between the MPI communicators for perturbations and kpts. Have you tried without paral_rf?