mods=(
  cuda/11.6
  intel/19.1.2
  mkl/2020.4
  impi/2019.8
  autoconf/2.69
  automake/1.15
  libtool/2.4.6
)


module purge
module load ${mods[@]}
LIB_PATH="${CUDA_HOME}/lib64"
export CUDA_ROOT=${CUDA_HOME}
export CUDA_LDFLAGS="-L${LIB_PATH} -lcuda -L${LIB_PATH} -lcudart -L${LIB_PATH} -lcublas"
export CUDA_CXXFLAGS="-I${CUDA_HOME}/include"

export LD_LIBRARY_PATH="${MKL_HOME}/lib/intel64_lin:${LD_LIBRARY_PATH}"

BLAS_STATIC_PATH="$MKL_HOME/lib/intel64/libmkl_intel_lp64.a"

ls ${LIB_PATH}/libcublas.so
ls ${LIB_PATH}/libcudart.so

cat <<EOF

////////////////////////////////////////////////////////////////////////////////
                                    info
////////////////////////////////////////////////////////////////////////////////


MKL_HOME         = $MKL_HOME
BLAS_STATIC_PATH = $BLAS_STATIC_PATH

CUDA_ROOT      = ${CUDA_HOME}
CUDA_LDFLAGS   = "-L${LIB_PATH} -lcuda -L${LIB_PATH} -lcudart -L${LIB_PATH} -lcublas"
CUDA_CXXFLAGS  = "-I${CUDA_HOME}/include"



Consider now runnng the following

../configure \\
  --enable-cuda \\
  --disable-slice \\
  --with-blas="-L\$MKL_HOME/lib/intel64/ -lmkl_intel_lp64 -mkl" \\
  CXX=mpiicpc \\
  CC=mpiicc \\
  MPICXX=mpiicpc


EOF
	

return