[Pw_forum] Possible issue with cp.x, ESSL FFT, and ultrasoft pseudopotentials on Blue Gene/Q

[Paolo Giannozzi]

This seems to happen only in the OpenMP version. Quick fix: use 
FFTW instead of ESSL for OpenMP parallelization or FFTs
(in make.sys, variables DFLAGS and FDFLAGS, set -D__LINUX_ESSL 
and -D__FFTW instead of -D__ESSL).

Paolo

On Thu, 2014-07-17 at 18:34 -0400, James Buchwald wrote:
> Hi all,
> 
> While using Quantum ESPRESSO 5.1 compiled with the ESSL libraries on an
> IBM Blue Gene/Q supercomputer, I've encountered a strange error when
> using ultrasoft pseudopotentials with cp.x.  While the calculation is
> setting up, an error in the ESSL FFT routine is thrown:
> 
>  > Wave Initialization: random initial wave-functions
>  > Occupation number from init
>  > nbnd =   512
>  > ...
>  > formf: eself=  2502.16598
>  > formf:     vps(g=0)=  -0.0003512     rhops(g=0)=  -0.0002271
>  > formf: sum_g vps(g)=  -3.0156511 sum_g rhops(g)=  -1.0523839
>  > formf:     vps(g=0)=  -0.0000199     rhops(g=0)=  -0.0000383
>  > formf: sum_g vps(g)=  -0.3905463 sum_g rhops(g)=  -0.7183677
>  > Delta V(G=0):   0.067259Ry,    1.830222eV
>  >
>  > DCFT : 2538-2014
>  > The routine must be initialized with the present value of (ARG NO. 9).
> 
> This error only occurs when using ultrasoft pseudopotentials;
> calculations with other pseudos complete successfully.  I encounter this
> both with my own input files and with example input files that have been
> modified to use ultrasoft pseudopotentials.  For example, here is an
> input file (with atomic positions omitted) from CP example 04 (128 water
> molecules) that has been modified to use ultrasoft pseudos; this is the
> input file that produced the above error.
> 
>  >  &control
>  >     title = ' Water 128 molecules ',
>  >     calculation = 'cp',
>  >     restart_mode = 'from_scratch', ! 'restart',
>  >     ndr = 50,
>  >     ndw = -1,
>  >     nstep  = 10,
>  >     iprint = 10,
>  >     isave  = 100,
>  >     tstress = .TRUE.,
>  >     tprnfor = .TRUE.,
>  >     dt    = 5.0d0,
>  >     etot_conv_thr = 1.d-8,
>  >     prefix = 'h2o',
>  >     pseudo_dir='/gpfs/u/home/QSFC/QSFCbcjm/barn-shared/uspp/',
>  > outdir='/gpfs/u/home/QSFC/QSFCbcjm/barn-shared/espresso-5.1/tempdir/',
>  >  /
>  >  &system
>  >     ibrav = 14,
>  >     celldm(1) = 18.65,
>  >     celldm(2) = 2.0,
>  >     celldm(3) = 2.0,
>  >     celldm(4) = 0.0,
>  >     celldm(5) = 0.0,
>  >     celldm(6) = 0.0,
>  >     nat = 384,
>  >     ntyp = 2,
>  >     nbnd = 512,
>  >     ecutwfc = 70.0,
>  >     ecfixed = 68.0,
>  >     qcutz = 68.0,
>  >     q2sigma = 8.0,
>  >     nr1b = 20,
>  >     nr2b = 20,
>  >     nr3b = 20,
>  >  /
>  >  &electrons
>  >     emass = 400.d0,
>  >     emass_cutoff = 2.5d0,
>  >     orthogonalization = 'ortho',
>  >     ortho_eps = 5.d-8,
>  >     ortho_max = 25,
>  >     electron_dynamics = 'sd',
>  >     electron_velocities = 'zero',
>  >     electron_temperature = 'not_controlled',
>  >  /
>  >  &ions
>  >     ion_dynamics = 'none',
>  >     ion_radius(1) = 0.8d0,
>  >     ion_radius(2) = 0.5d0,
>  > !    ion_velocities = 'zero',
>  >     ion_temperature = 'not_controlled'
>  >  /
>  >  &cell
>  >     cell_dynamics = 'none',
>  >     cell_velocities = 'zero',
>  >     press = 0.0d0,
>  >     wmass = 70000.0d0
>  >  /
>  > ATOMIC_SPECIES
>  >  O  16.0d0   O.pbe-van_bm.upf
>  >  H  1.0079d0 H.pbe-van_bm.upf
>  > ATOMIC_POSITIONS (crystal)
> 
> Any help or advice would be appreciated!
> 
> Best,
> James
> 

-- 
 Paolo Giannozzi, Dept. Chemistry&Physics&Environment, 
 Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
 Phone +39-0432-558216, fax +39-0432-558222