Installation of AMBER
Step 1: Setting Environment Variable
what package/library you need for compile AMBER is GNU compiler, FFTW, MPI or OpenMPI, and CUDA (only for using GPU)
export AMBERHOME=$HOME/amber
export CUDA_HOME=$CRAY_CUDATOOLKIT_DIR
export CRAYPE_LINK_TYPE=dynamic
export CRAY_ADD_RPATH=yes
Then you are supposed ready to compile AMBER. Below are different commands which would compile AMBER as your desired runtime version.
Step 2: Compile AMBER
# Compile AMBER as serial CPU version
make clean
./configure -nomtkpp -crayxt5 -noX11 --with-python $PYTHON --with-netcdf $NETCDF_DIR gnu
make install
# Compile AMBER as parallel CPU version
make clean
./configure -nomtkpp -crayxt5 -noX11 --with-python $PYTHON --with-netcdf $NETCDF_DIR -mpi gnu
make install
# Compile AMBER as serial GPU version
make clean
find . -name "*.a" -exec rm {} \; find . -name "*.so" -exec rm {} \; find . -name "*.so.*" -exec rm {} \; find . -name netcdf.mod -exec rm {} \; export PYTHON=`which python`
./configure -nomtkpp -crayxt5 -noX11 --with-python $PYTHON --with-netcdf $NETCDF_DIR -cuda gnu
make install
# Compile AMBER as parallel GPU version
make clean
./configure -nomtkpp -crayxt5 -noX11 --with-python $PYTHON --with-netcdf $NETCDF_DIR -cuda -mpi gnu
make install
The AMBER executable file is at $AMBERHOME/bin directory.
Step 3: Test AMBER
cd $AMBERHOME/test/ubiquitine
$AMBERHOME/bin/sander.MPI -O -i mdin -p prmtop -c inpcrd -o job.out
Rangsiman Ketkaew