[Pw_forum] AIX + XML files

[Axel Kohlmeyer]

On Sun, 26 Nov 2006, Gianni Profeta wrote:

GP> On Sat, 25 Nov 2006, Axel Kohlmeyer wrote:
GP> 
GP> > CP>       from write_rho_xml : error #         1
GP> > CP>       cannot opentmp/Al.save/charge-density.xml file for writing
GP> > CP> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
GP> >
GP> > please check your input file. it looks as if you used
GP> > 'tmp' instead of '/tmp'. all QE codes will fail if you
GP> > try to write to a non-existing directory...
GP> >
GP> > axel.
GP> 
GP> Dear Alex,
GP> I already did. I used :    outdir='tmp/'
GP> and indeed all the previous versions (<= 3.0) work for the same  input 
GP> file.

ok. that is indeed a hint, that you are seeing something 
related to iotk.

please have a look at your make.sys file and compare the settings
that configure has picked up with the following (most important are
F90 and MPIF90, this is for parallel compilation...). there are
some subtle semantics issue, that strike only the ibm compilers
which do not seem to automatically switch to f95 mode on .f90 file
name extensions (you _have_ to use xlf95 or mpxlf95... :-( ).

# If the mass library is available, use -D__MASS

DFLAGS     = -D__AIX,-D__XLF,-D__ESSL,-D__MASS,-D__PARA,-D__MPI
CPPFLAGS   = -D__AIX -D__XLF -D__ESSL -D__MASS -D__PARA -D__MPI

# Fortran and C compilers 
CC           = mpcc
MPICC        = mpcc
F90          = mpxlf95 
MPIF90       = mpxlf95
F77          = mpxlf 
MPIF77       = mpxlf

# Fortran flags

FFLAGS       = -qalias=noaryovrlp:nointptr -I../include \
               -O3 -qstrict -qarch=auto -qtune=auto -qdpc -Q -qessl
FFLAGS_NOOPT = -O0
F90FLAGS     = $(FFLAGS) -qfree=f90 -qsuffix=cpp=f90 -WF,$(DFLAGS) \
               -I../Modules -I../PW -I../PH -I../iotk/src
CFLAGS       = -I../include $(CPPFLAGS) -O2

# For essl and mass libraries
 
LIBS         = -lessl -lmassvp4 -lmass #-L/usr/local/lib -llapack

# Linker
LD           = $(MPIF90)


GP> 
GP> I think that  if 'outdir' does not exist is checked before the writing, 
GP> and the error message in that case is:
GP> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
GP>       from outdir:  : error #         1
GP>       tmp/ non existent or non writable
GP> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
GP> 
GP> just after the execution starts.
GP> 
GP> While in my case, the code arrives at the beginning of the first iteration 
GP> and then stops.
GP> In fact, it can writes into the tmp/ directory :
GP> 
GP> >ls -ltr tmp/
GP> -rw-r--r--   1 gianni   staff             0 Nov 26 09:07 Al.wfc1
GP> drwxr-xr-x   2 gianni   staff           256 Nov 26 09:07 Al.save
GP> -rw-r--r--   1 gianni   staff             0 Nov 26 09:07 Al.igk
GP> 
GP> but the files and directory (Al.save) are empty.

ok. this was just a wild guess. if you intended to write to tmp/ you
are correct. people usually do not use a path like that for scratch 
files, as it is generally NFS mounted, and thus lead to bad i/o 
performances...


ciao,
   axel.

GP> 
GP> Thank you.
GP> Gianni
GP> _______________________________________________
GP> Pw_forum mailing list
GP> Pw_forum at pwscf.org
GP> http://www.democritos.it/mailman/listinfo/pw_forum
GP> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.