To compile RMPCDMD, the following software is required

Other utilities, mostly Fortran code, will be downloaded automatically as part of the installation process.

RMPCDMD has been developed on Linux, using both the GFortran and Intel Fortran compilers. Compilation under OS X has been successfully achieved by users of RMPCMD. The Windows platform is currently not tested.

If you encounter a configuration issue, it is often useful to remove all CMake data by executing:

rm -r CMake*

in the build directory (not in the source directory).

After successfully building the program, copy the file rmpcdmd in a location where executables are found (i.e. $HOME/.local/bin or $HOME/bin for instance). This file provides a unique interface to RMPCDMD.

To analyze simulation data, RMPCDMD provides the command rmpcdmd plot (see Run the code) that requires Python and a number of packages (see below).


Once CMake has been executed, the RMPCDMD directory cannot be moved. If it is moved, the complete configuration procedure has to be restarted and the rmpcdmd program (in $HOME/.local/bin or $HOME/bin) must be replaced by the newly created version.


The rmpcdmd program does not contain the simulations programs themselves and relies on their presence at the location of compilation.

Building on Linux

On a Debian distribution (or derivative, e.g. Ubuntu), as root:

apt-get install gfortran libhdf5 cmake git

will install the dependencies. (See install_hdf5 for Ubuntu 14.04 or problematic HDF5 installs).

At the command line, execute the following:

git clone
git submodule init
git submodule update
mkdir build
cd build
cmake ..
make VERBOSE=1

This will first fetch RMPCDMD, then the related modules ParseText, fortran_tester, fortran_h5md and random_module. The compilation is prepared by the cmake command and executed by the make command.

If your HDF5 installation is properly setup, CMake should find it automatically. If this is not the case, you may define the environment variable HDF5_ROOT=/path/to/hdf5 before invoking cmake (see the OS X installation notes for an example).


When using compilers that are not the default one of the Linux distribution, such as the Intel compilers, the following settings must be used (change the compiler name if needed) before running cmake and before possibly building HDF5:

export CC=icc
export FC=ifort

Building on OS X

The first step is to install a development environment, and thus the XCode software from Apple (we have tested RMPCDMD with XCode 7.3.1 on OS X El Capitan).

We used MacPorts for gcc, git, cmake and make. Install MacPorts (a GUI installer is available at and then the dependencies:

sudo port install gcc-5 cmake git

The rest is similar to Linux except that the name of the compiler has to be specified manually at the command-line to prevent the automatic selection of Apple’s provided compiler. Also, HDF5 is built locally:

git clone
git submodule init
git submodule update
export CC=gcc-mp-5
export FC=gfortran-mp-5
mkdir build
cd build
HDF5_ROOT=../_hdf5-1.8.17 cmake ..

The compiler names given here may vary depending on your setup.

Building on Windows

Installation on Windows is not tested at this time. The major difficulty is likely related to a functional installation of HDF5 for Fortran under Windows, that is only provided for the Intel compiler. Having no Windows computer with the Intel Fortran compiler at our disposal, user feedback is welcome.


On OS X or with older Linux systems (such as Ubuntu 14.04 still in wide usage), it is necessary to build HDF5 to enable the Fortran 2003 interface.

A script is provided to download and build HDF5 that has been tested on OS X and on Ubuntu. It must be run from the main RMPCDMD directory:


with compiler definitions CC and FC set when using OS X or a compiler that is not the default one for the computer, such as icc and ifort. This script must be run only once and the environment variable HDF5_ROOT must be set when invoking cmake (also see the OS X installation notes).

The script downloads HDF5 1.8.17 and installs it under _hdf5-1.8.17. You can remove the directory hdf5-1.8.17 (no leading underscore) after the execution of the script.


Several analysis scripts in the experiments/ directory and a command-line tool rmpcdmd plot are provided. They all rely on the Python programming language and the following Python packages:

Optionally, you may wish to install Mayavi for the 3D visualization of the dimer simulation.

Installing those tools under Linux is straightforward:

sudo apt-get install python-numpy python-scipy python-matplotlib python-h5py

for Debian-based systems or:

sudo yum install numpy scipy h5py python-matplotlib

for Red-Hat based systems.

On OS X, we recommend to use a Python “super package” such as Enthought Canopy or Anaconda from Continuum that bundle the required software.