BRAHMS is a simulation execution engine which was written by Ben Michinson. Amoungst other things, it's often used in SpineML_2_BRAHMS to provide a simulation back-end for SpineML models.
This, now official, version of BRAHMS was branched at version 0.7.3 of the code from brahms.sourceforge.net.
You can read all about the original at:
http://brahms.sourceforge.net/docs/What%20is%20BRAHMS.html
The original code is also available at that site.
The major change between this version and BRAHMS 0.7.3 is that the
requirement to set SYSTEMML_INSTALL_PATH has been removed. Instead, when
the software is compiled, a directory for the "installed SystemML Namespace"
is set - that might be /usr/local/var/SystemML/Namespace. That becomes
the default BRAHMS Namespace. It's in a var directory as you may decide
to install your own components in there.
You can also run BRAHMS without the progress window with the --nogui option. This can be useful if you need to run hundreds of concurrent BRAHMS instances on an HPC system and you forgot to log in with a non-X windows connection.
I may re-instate SYSTEMML_INSTALL_PATH such that you can have a second
installed Namespace, perhaps at $HOME/SystemML/Namespace which is referred
to with SYSTEMML_INSTALL_PATH.
I may implement a scheme for saving the delayed values stored in connections which have some specified lag.
This project uses cmake for a cross-platform friendly build process. cmake is a good choice of build system, because it is familiar to many developers.
You'll need to obtain the correct dependencies to build BRAHMS. On Linux that means a compiler, libXv, libXaw and WX windows. For example, on (current 2015) Ubuntu or Debian systems you'll want to do:
sudo apt-get install build-essential libxaw7-dev libxv-dev zlib1g-dev
These libraries are used to draw the progress box which shows while BRAHMS is running.
For Windows, you'll reportedly need GnuWin32, but this build has not yet been tested.
For optional features (MPI channel, Python and MATLAB bindings and one WXWidgets-dependent component) you'll need mpich2, python development libraries, WXwidgets and a MATLAB installation. The first three can be obtained via apt-get:
sudo apt-get install mpich2 python-dev libwxgtk2.8-dev
The recommended method for building is to create a separate build folder as follows:
mkdir build
cd build
cmake -DCMAKE_INSTALL_PREFIX=/usr/local .. # or cmake-gui ..
make -j4 # or however many processor cores you have
sudo make install
Here are the options you can set. Shown here are the defaults (at time of writing):
Adding/removing bindings:
cmake -DCOMPILE_MATLAB_BINDING=OFF -DCOMPILE_PYTHON_BINDING=ON ..
MPICH2:
cmake -DCOMPILE_WITH_MPICH2=ON ..
Compiling a component which requires WX windows:
cmake -DCOMPILE_WX_COMPONENT=OFF ..
Compiling with a specific installation prefix:
cmake -DCMAKE_INSTALL_PREFIX=/usr/local ..
-- Seb James, Feb 2015.