[CP2K:167] Re: about output handling in cp2k

[Teodoro Laino]

This is just a follow-up of the few messages between me and Axel  
about the problem on the CRAY experienced running the input
cp2k/tests/QS/H2O-32.inp  (maybe interesting for people following  
this discussion):

> use matthias H2O-32.inp. it dies for me after completing 8 scf cycles
> of the first MD step running over 8 PEs.

On our CRAY it's working both with
pgi-6.1.4 and pgi-6.2.5 and pgi-7.0.4

it works also with

pathscale-3.0

NO way to get cp2k working with gfortran..

This is the arch file I'm using for pgi:
-----------------------
# The following settings worked for a Cray XT3 using PGI 6.0.5 and  
ACML 2.6
# Note: PGI 5.2.* and 6.0.1-4 are not working due to compiler bugs.
# Note: The -g flag is needed for unknown reasons.
PERL     = perl
CC       = gcc
CPP      = cpp
FC       = ftn -Mfree
LD       = ftn
AR       = ar -r
DFLAGS   = -D__XT3 -D__FFTSG\
            -D__parallel -D__BLACS -D__SCALAPACK
CPPFLAGS = -traditional -C $(DFLAGS) -P
FCFLAGS  = -target=catamount -O2 -Mcache_align
FCFLAGS2 = -target=catamount -O0 -Mcache_align
LDFLAGS  = $(FCFLAGS)
LIBS     = /users/teo/acml3.0.0/pgi64/lib/libacml.a

OBJECTS_ARCHITECTURE = machine_xt3.o


qs_oce_methods.o: qs_oce_methods.f90
         $(FC) -c $(FCFLAGS2) $<
qs_oce_types.o: qs_oce_types.f90
         $(FC) -c $(FCFLAGS2) $<
hartree_local_2centers.o: hartree_local_2centers.f90
         $(FC) -c $(FCFLAGS2) $<
hartree_local_3centers.o: hartree_local_3centers.f90
         $(FC) -c $(FCFLAGS2) $<
qs_ks_atom.o: qs_ks_atom.f90
         $(FC) -c $(FCFLAGS2) $<
qmmm_gpw_energy.o: qmmm_gpw_energy.f90
         $(FC) -c $(FCFLAGS2) $<
---------------------------

Are you running the 8 procs on 4 nodes  (is it a dual core?)..
Could it be a memory problem? any message from the error file?
If it's a dual core run try to run 8 (1 per core) forcing yod with - 
SN just to check
if it's really a memory problem..

Teo