[Pw_forum] Problem with a run on IBM

[Giovanni Cantele]

I'm trying to do a very simple run on an IBM sp3 machine. I compute just 
the total energy for a single H atom, using the last version of PWSCF.
The input file reads as follows:


  &CONTROL
         calculation='scf'
         title='H'
         outdir='./'
         prefix='H'
         pseudo_dir='/u/cantele/RUNS/pseudo/'
  /
  &SYSTEM
         ibrav=1
         celldm(1)=15.0
         nat=1
         ntyp=1
         ecutwfc=15.0
         occupations='smearing'
         degauss=0.005
  /
  &ELECTRONS
         electron_maxstep=60
         conv_thr=1.0d-6
         mixing_beta=0.7d0
  /
ATOMIC_SPECIES
         H      1.00794    H.pz-rrkjus.UPF
ATOMIC_POSITIONS { bohr }
          H    0.0    0.0    0.0
K_POINTS { gamma }



This run seems to work fine. Neverthless I encountered problems when 
using  H.pz-vbc.UPF or the old H.vbc pseudopotential. In the first case
the end of the output file is



      G cutoff =  341.9590  (  26529 G-vectors)     FFT grid: ( 40, 40, 40)

   +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++
     Fatal error in routine `allocate_real_4d': Memory allocation failed
   +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++
   Called by
   +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++ +++



and the run error message

ERROR: 0032-184 MPI was not finalized  in routine unknown




In the second case there is no error message within the output file 
which stops at


Reading pseudopotential file in UPF format


and the run error message is

Error encountered while attempting to allocate a data object.


The three runs all work on an alpha machine. Perhaps there should be
some error in compiling on sp. Does anyone know what is the problem?

Thanks,
Giovanni