problem of recompiling cp2k

[Axel]

On Aug 5, 5:06 pm, Jun <chen... at googlemail.com> wrote:
> Hi,
>
> I have compiled cp2k using openmpi successfully. However, the
> executable cannot pass any single test.
> It complained unknown keywards in input files:
>
>             2 section_vals_parse
>             1 create_cp2k_input_reading
>  **** file created by ****
>  *
>  *** 17:53:38 ERRORL2 in input_parsing:section_vals_parse err=-300
> found  ***
>  *** an unknown keyword **** in section
> __ROOT__                           ***
>
> But I am very sure there is no problem with the input file. I also
> tried the test input files in cp2k/tests.
> It should be the problem in executable.

which cp2k tests input fails?

if it is a problem with the executable, then i has to be due to the
compiler misompiling it. the "safe" fallback is usually g95. i always
compile both, intel and g95 versions of the executables, so that i
can conveniently crosscheck if it is a compiler problem or a source
code problem.

cheers,
   axel.

>
> I am using
> intel/cce/11.0.081 (11.1 doesn't work with cp2k, I saw a post in intel
> forum talking about this problem)
> intel/fce/11.0.081
> intel/cmkl/10.2.1.017
> fftw/intel/64/3.2.1
> openmpi/1.3.1
>
> Here is my arch file
> CC       = icc
> CPP      =
> FC       = mpif90
> LD       = mpif90
> AR       = ar -r
> DFLAGS   = -D__INTEL -D__FFTSG -D__parallel -D__BLACS -D__SCALAPACK -
> D__FFTW3
> CPPFLAGS =
> FCFLAGS  = $(DFLAGS) -O3 -xT -funroll-loops -fpp -free -i_dynamic
> LDFLAGS  = $(FCFLAGS)
> INTEL_LIB = /usr/local/Cluster-Apps/intel/cmkl/10.2.1.017/lib/em64t
> FFTW_LIB = /usr/local/Cluster-Apps/fftw/intel/64/3.2.1/lib
> LIBS     = -L$(INTEL_LIB)              \
>            -lmkl_scalapack_lp64        \
>            -lmkl_blacs_openmpi_lp64    \
>            -lmkl_lapack                \
>            -lmkl_intel_lp64 -lmkl_intel_thread -lmkl_lapack -lmkl_core
> \
>            -lguide -lpthread           \
>            -L$(FFTW_LIB)               \
>            -lfftw3
> OBJECTS_ARCHITECTURE = machine_intel.o
>
> Thanks in advance.
>
> Regards,
> Jun
>
> On Aug 3, 4:44 pm, Jun <chen... at googlemail.com> wrote:
>
> > Thanks a lot, Teo.
>
> > Jun
>
> > On Aug 3, 3:14 pm, Teodoro Laino <teodor... at gmail.com> wrote:
>
> > > Dear Jun,
>
> > > with some recent changes to mpi, the only way to compile cp2k is with
> > > MPI 2.0.
> > > There was already a message on this topic from Axel:
>
> > >http://groups.google.com/group/cp2k/browse_thread/thread/7da466187cd1...
>
> > > Teo
>
> > > Jun wrote:
> > > > Dear all,
>
> > > > I am trying to update cp2k to the latest version, and recompiling it.
> > > > But I failed using the same arch file, compiler (fce/11.0.074) and
> > > > libraries (cmkl/10.1.0.015). The error appears at compiling
> > > > message_passing.F:
> > > > *************************************************************************
> > > > mpif90 -c -D__INTEL -D__FFTSG -D__parallel -D__BLACS -D__SCALAPACK -
> > > > D__FFTW3 -I//usr/local/Cluster-Apps/intel/cmkl/10.2.1.017/include/ -I//
> > > > usr/local/Cluster-Apps/fftw/intel/64/3.2/include/ -O3 -xW -heap-arrays
> > > > 64 -funroll-loops -fpp -free -D__COMPILE_ARCH="\"Linux-x86-64-intel
> > > > \""  /home/jc590/cp2k/makefiles/../src/message_passing.F
> > > > /home/jc590/cp2k/makefiles/../src/reference_manager.F(214): (col. 8)
> > > > remark: LOOP WAS VECTORIZED.
> > > > /home/jc590/cp2k/makefiles/../src/reference_manager.F(217): (col. 11)
> > > > remark: LOOP WAS VECTORIZED.
> > > > /home/jc590/cp2k/makefiles/../src/reference_manager.F(222): (col. 8)
> > > > remark: LOOP WAS VECTORIZED.
> > > > mpif90 -c -D__INTEL -D__FFTSG -D__parallel -D__BLACS -D__SCALAPACK -
> > > > D__FFTW3 -I//usr/local/Cluster-Apps/intel/cmkl/10.2.1.017/include/ -I//
> > > > usr/local/Cluster-Apps/fftw/intel/64/3.2/include/ -O3 -xW -heap-arrays
> > > > 64 -funroll-loops -fpp -free -D__COMPILE_ARCH="\"Linux-x86-64-intel
> > > > \""  /home/jc590/cp2k/makefiles/../src/cp_files.F
> > > > /home/jc590/cp2k/makefiles/../src/message_passing.F(5111): error
> > > > #6404: This name does not have a type, and must have an explicit
> > > > type.   [MPI_LOCK_EXCLUSIVE]
> > > >     exclusivity = MPI_LOCK_EXCLUSIVE
> > > > ------------------^
> > > > /home/jc590/cp2k/makefiles/../src/message_passing.F(5113): error
> > > > #6404: This name does not have a type, and must have an explicit
> > > > type.   [MPI_LOCK_SHARED]
> > > >        IF (shared) exclusivity = MPI_LOCK_SHARED
> > > > ---------------------------------^
> > > > /home/jc590/cp2k/makefiles/../src/message_passing.F(5117): error
> > > > #6404: This name does not have a type, and must have an explicit
> > > > type.   [MPI_MODE_NOCHECK]
> > > >        IF (nocheck) assertion = MPI_MODE_NOCHECK
> > > > --------------------------------^
> > > > /home/jc590/cp2k/makefiles/../src/message_passing.F(5193): error
> > > > #6404: This name does not have a type, and must have an explicit
> > > > type.   [MPI_MODE_NOSTORE]
> > > >        IF (nothing_stored) assertion = assertion + MPI_MODE_NOSTORE
> > > > ---------------------------------------------------^
> > > > /home/jc590/cp2k/makefiles/../src/message_passing.F(5196): error
> > > > #6404: This name does not have a type, and must have an explicit
> > > > type.   [MPI_MODE_NOPUT]
> > > >        IF (no_put_follows) assertion = assertion + MPI_MODE_NOPUT
> > > > ---------------------------------------------------^
> > > > /home/jc590/cp2k/makefiles/../src/message_passing.F(5199): error
> > > > #6404: This name does not have a type, and must have an explicit
> > > > type.   [MPI_MODE_NOPRECEDE]
> > > >        IF (no_pre_access) assertion = assertion + MPI_MODE_NOPRECEDE
> > > > --------------------------------------------------^
> > > > /home/jc590/cp2k/makefiles/../src/message_passing.F(5202): error
> > > > #6404: This name does not have a type, and must have an explicit
> > > > type.   [MPI_MODE_NOSUCCEED]
> > > >        IF (no_post_access) assertion = assertion + MPI_MODE_NOSUCCEED
> > > > ---------------------------------------------------^
> > > > compilation aborted for /home/jc590/cp2k/makefiles/../src/
> > > > message_passing.F (code 1)
> > > > make[1]: *** [message_passing.o] Error 1
> > > > make[1]: *** Waiting for unfinished jobs....
> > > > *************************************************************************
>
> > > > my arch file is
>
> > > > CC       = icc
> > > > CPP      =
> > > > FC       = mpif90
> > > > LD       = mpif90
> > > > AR       = ar -r
> > > > DFLAGS   = -D__INTEL -D__FFTSG -D__parallel -D__BLACS -D__SCALAPACK -
> > > > D__FFTW3
> > > > CPPFLAGS =
> > > > FCFLAGS  = $(DFLAGS) -O3 -xT -funroll-loops -fpp -free
> > > > LDFLAGS  = $(FCFLAGS)
> > > > INTEL_LIB = /usr/local/Cluster-Apps/intel/cmkl/10.1.0.015/lib/em64t
> > > > FFTW_LIB = /usr/local/Cluster-Apps/fftw/intel/64/3.2/lib
> > > > LIBS     = $(INTEL_LIB)/libmkl_scalapack.a \
> > > >            $(INTEL_LIB)/libmkl_blacs_lp64.a \
> > > >            $(INTEL_LIB)/libmkl_lapack.a \
> > > >            $(INTEL_LIB)/libmkl_em64t.a \
> > > >            $(INTEL_LIB)/libguide.a \
> > > >            $(FFTW_LIB)/libfftw3.a
>
> > > > OBJECTS_ARCHITECTURE = machine_intel.o
> > > > *************************************************************************
>
> > > > I noticed message_passing.F was changed after my last successful
> > > > compilation. Not sure if it is relevant. Has anyone seen similar
> > > > errors before? Any suggestion will be appreciated.
>
> > > > Regards,
> > > > Jun