Linking Problems when compiling NWChem 6.3

#!/bin/bash
export NWCHEM_TOP=$HOME/Programs/nwchem-6.3-src.2013-05-17
export NWCHEM_TARGET=LINUX64
export USE_MPI=y
export USE_MPIF=y
export USE_MPIF4=y
export MPI_LOC=$HOME/Programs/OpenMPI
export LIBMPI="-lmpi_f90 -lmpi_f77 -lmpi -lpthread"
export MPI_LIB=$MPI_LOC/lib
export MPI_INCLUDE=$MPI_LOC/include
export TCE_CUDA=Y
export CUDA_LIBS="-L/usr/local/cuda/lib64 -L/usr/local/cuda/lib -lcudart"
export CUDA_FLAGS="-arch sm_20 "
export CUDA_INCLUDE="-I. -I/usr/local/cuda/include"
export LARGE_FILES=TRUE
export NWCHEM_MODULES=all
export MKLROOT=/opt/intel/mkl/lib/intel64
export BLASOPT="-L$MKLROOT -lmkl_scalapack_ilp64 -lmkl_cdft_core -lmkl_gf_ilp64 -lmkl_sequential -lmkl_core -lmkl_blacs_intelmpi_ilp64 -ldl -lpthread -lm"
make clean
make FC=gfortran FFLAGS="-fdefault-integer-8 -m64" CUDA=nvcc nwchem_config
make FC=gfortran FFLAGS="-fdefault-integer-8 -m64" CUDA=nvcc HAS_BLAS=yes

Doudou,
Have you tried to do a regular NWChem installed without any Cuda bit? Was it successful?
Please do not set the FFLAGS variable since the NWChem makefle structure takes care of it.

As far as Cuda is concerned, I have a few questions for you:

1) Could you please show me the files that appear in you CUDA Library directories. What CUDA version have you installed?
2) Do you have g++ installed?
3) What kind of GPU Hardware do you have

Doudou,
I have managed to reproduce the first four linking failures you reported.
The fix is to switch to ARMCI_NETWORK=SOCKETS (instead of ARMCI_NETWORK=MPI-TS that is used by default)

To switch:

export ARMCI_NETWORK=SOCKETS
cd $NWCHEM_TOP/src/tools
rm -rf build install
make
cd ..
make link


Please let me know if this fixes your first four linking errors

I did some research on the other errors, there is one in particular the undefined reference to 'std::_Rb_tree_increment(std::_Rb_tree_node_base const*)' that has to do with using gcc libraries with g++ or vice versa. I am not sure if this helps in this case. Thanks again!

Have tried to compile and link the sample programs from the Cuda SDK to test you software environment?

Edo

Doudou,

Can you post the following.

1. Output of make inside src/tce/ccsd_t/

2. Can you try linking with -lstdc++

Kiran

Doudou,
Since we are at a loss at reproducing your linking error, could you please post as many details as possible about your compialtion environment?
For example, please let us know
1) Linux distribution & version
2) gcc, g++ and gfortran versions and origin (i.e. how where they installed)
3) Cuda SDK version and it was installed.

Thanks, Edo