Building CN24

Dependencies

CN24 uses the CMake cross-platform build system. You need at least version 2.8 to generate the build files. The following compilers are supported for building CN24:

• GCC >= 4.9 on Linux (or >= 4.7 with Boost >= 1.54)
• Clang >= 3.5 on OS X
• Visual Studio >= 2013 on Windows

Older versions will probably work as long as they support the C++11 features used by CN24. All other dependencies are optional. Optional dependencies include:

• libjpeg and libpng to read .jpg and .png files
• Intel MKL, AMD ACML or ATLAS for faster calculations
• OpenCL and clBLAS for GPU acceleration (either both or none!)
• GTK+ 3 for GUI utilities

Installing clBLAS

Consider installing clBLAS for high performance training and prediction. Installation is simple:

• Linux: Install boost using your distribution's package manager, then run the included ci/install_clblas.sh script.
• Mac OS X: Install clBLAS using the homebrew package manager: brew install clblas
• Windows: Install a binary release from the clBLAS GitHub releases page.

Note: for all systems except for Mac OS X you have to install the correct OpenCL libraries for your GPU beforehand. Windows OpenCL implementations are part of AMD APP SDK and nVidia CUDA Toolkit respectively.

Note: clBLAS is currently broken for nVidia users, please use clBLAS v2.6 for the time being.

Building CN24

First, make sure you have all the required dependencies. Then, clone the CN24 repository:

git clone https://github.com/cvjena/cn24.git

Create a build directory, switch to it and run CMake:

mkdir build
cd build
cmake path/to/cn24

Take a moment and run ccmake (or cmake-gui) to see all the available build options. We highly recommend selecting at least the libpng and/or libjpeg options:

ccmake .

Run your preferred build tool, for example:

make

That's it, you're done!