AMBER16+CUDA Installation in Ubuntu 18.04 (BIONIC)

---

Author: Senthilkumar Kailasam

Date: Feb 2018

• Here I will go through the steps involved in installing amber16 + Ambertools 17 in Ubuntu 18.04 LTS with cuda 9.0 and nvidia-driver-390

Note:If you don't have a GPU card you can skip the steps involving CUDA and NVIDIA-DRIVER

1. Install all prerequisites/dependency

• Assuming its a fresh installation of Ubuntu 18.04 LTS

### commands to install packages in ubuntu box as a root user.

apt-get install aptitude
aptitude install vim csh flex patch gfortran g++ make xorg-dev bison libbz2-dev 

### Parallel MPI packages 

aptitude install openmpi-bin libopenmpi-dev 

### Python packages 

aptitude install python-numpy python-matplotlib python-scipy ipython  python-notebook cython python-dev python-pip 

2. GNU GCC 6 Installation

• Ubuntu 18.04 LTS comes with GCC 7 and AMBER16 requires GCC 6.

### GCC6 installation
aptitude install gcc-6 g++-6 gfortran-6

#### create Symbolic Links in /usr/bin
rm -f gcc gcc-ar gcc-nm gcc-ranlin g++ gfortran
ln -s gcc-6 gcc
ln -s gcc-ar-6 gcc-ar
ln -s gcc-nm-6 gcc-nm
ln -s gcc-ranlib-6 gcc-ranlib
ln -s g++-6 g++
ln -s /usr/bin/gfortran-6 /usr/bin/gfortran 

3. NETCDF installation

• Download parallel-netcdf-1.9.0

### Install netcdf

## go to netcdf folder followed by make and make install
cd parallel-netcdf-1.9.0/ 
./configure

make 

make install 

4. CUDA 9.0 Installation

• Assuming that the NVIDIA DRIVER installation went well! If not see that instruction at Step 6
• Download cuda_9.0.176_384.81_linux.run from NVIDIA website.

### On the terminal , install cuda
wget https://developer.nvidia.com/compute/cuda/9.0/Prod/local_installers/cuda_9.0.176_384.81_linux-run

./cuda_9.0.176_384.81_linux.run 

### add lines to ld.so.conf

echo "include /usr/local/cuda-9.0/lib64" >> /etc/ld.so.conf 

### run ldconfig

ldconfig 

### Add PATH and environment updates to ~/.bashrc file and source

export PATH="/usr/software/anaconda2/bin:$PATH"
test -f /usr/software/amber16/amber.sh && source /usr/software/amber16/amber.sh

export PATH=/usr/local/cuda-9.0/bin${PATH:+:${PATH}}export 
LD_LIBRARY_PATH=/usr/local/cuda/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}} 

source ~/.bashrc 

5. Compiling AMBER 16 and AMBERTOOLS 17

• Download AMBER with license purchase at ambermd.org
• Download AmberTools17 from ambermd.org

### Steps on the terminal

### Steps to compile the serial verison assuming my files are in tmp folder.

tar xvvf /tmp/Amber16.tar.bz2 

tar xvvf /tmp/AmberTools17.tar.bz2 

cd /usr/software/amber16 

./configure gnu 

make install 

### Steps to compile the parallel verison with NETCDF capabilities

./configure -mpi --with-pnetcdf /usr/software/parallel-netcdf gnu 

make install 

#### Steps to compile serial cuda enabled amber 

./configure -cuda gnu 
make install 
    
#### Steps to compile parallel+netcdf+cuda enabled amber 
./configure -mpi -cuda --with-pnetcdf  /usr/soft/parallel-netcdf gnu 

make install 

6. NVIDIA DRIVER INSTALLATION

I followed the instruction given at https://www.mvps.net/docs/install-nvidia-drivers-ubuntu-18-04-lts-bionic-beaver-linux/

### Clean the system of other Nvidia drivers

aptitude purge nvidia*

### check the latest driver version for our Nvidia GPU

using link https://www.nvidia.com/object/unix.html
    
### Add the Nvidia graphic card PPA

add-apt-repository ppa:graphics-drivers
    
### Prepare the system for the installation

aptitude update

### Install the Nvidia GPU driver

aptitude install nvidia-driver-390

reboot

#### Verify Nvidia Driver installation

lsmod | grep nvidia
or
nvidia-smi