Installation in Ubuntu¶
In the following a quick step-by-step guide how to download and install HALMD in Ubuntu will be given.
Ubuntu 14.04 LTS¶
Prepare installation¶
First, make sure you have all required packages installed:
sudo apt-get install git build-essential libluajit-5.1-dev zlib1g-dev wget nvidia-cuda-toolkit
Clone the HALMD source code repository:
git clone --recursive http://git.halmd.org/halmd.git
Build and install prerequisites¶
Then you will have to build boost from source using examples/packages.mk
:
mkdir halmd-prerequisites
cd halmd-prerequisites
nice make CONCURRENCY_LEVEL=4 -f ../halmd/examples/packages.mk install-boost install-hdf5 install-cmake
This step is required as using the according packages from the Ubuntu repository will not work—all these packages need special build options not provided in the official packages. Note that this installation will require some time but you need to do this only once. A more detailed description of the package installation can be found in Automatic installation.
After executing the above commands, the necessary packages will be installed in ~/opt
.
In order to make these packages available for the subsequent build tools, run
source <(make -f ../halmd/examples/packages.mk env-boost env-hdf5 env-cmake)
Alternatively, you may append the output to your ~/.bashrc
. You can verify
that everything went well by running cmake --version
which should output
something like cmake version 2.8.12.1 with native CUDA support
(the
important part is with native CUDA support
).
Build and install HALMD¶
Now, we can start building HALMD. First, create a build directory (this can be anywhere, for convenience we create it in our home directory) and then run cmake to generate the necessary Makefiles
mkdir ~/halmd-build && cd ~/halmd-build
cmake ~/halmd -DCMAKE_INSTALL_DIRECTORY=~/opt/halmd/
There may be warnings now about missing packages (e.g. Sphinx) but this is not
essential now as it is only required to build the manual page. If you are
building with CUDA support, make sure that the CUDA compiler has been detected
and works. If there was a problem and you were able to fix this, it may be
necessary to remove the build directory completely and rerun cmake ~/halmd
afterwards.
You are now ready to build HALMD. Execute
nice make -j4
and the build process will start.
Note
The build process is very memory hungry, consider reducing the number of
parallel builds to a lower number (i.e. use -j2
instead of -j4
if
you want 2 parallel builds) if you experience problems.
If successful, you can run ./halmd/halmd --version
from the build directory
and this should give you simple version information about HALMD.
If you want to install HALMD, simply run make install
from the build
directory. In order to be able to run halmd from everywhere, run
echo "export PATH=\"${HOME}/opt/halmd/bin\${PATH+:\$PATH}\"" >> ~/.bashrc
and log out and in again.
You can test now a simple example by running
halmd ~/opt/halmd/share/doc/halmd/examples/liquid/lennard_jones_equilibration.lua -v
You may now clean-up the build directories halmd-prerequisites
and halmd-build
.