NUMCXX Installation

The library has been tested under Linux, MacOSX, Win10/Cygwin

Prerequisites

Installation of the python components is currently not necessary.

Linux

The following packages should be installed using the Linux package manager (from system to system, they have different names)

gcc
g++
cmake 
blas
lapack 
suitesparse 
python
python-devel
python-numpy
python-matplotlib
swig

Mac

We need the same packages as on Linux.

• Install Xcode from the App-Store
• Trigger installaion of command line developer tools in the terminal via

$ gcc

• A dialogue window should pop up, click on install

$ xcode-select -p /Library/Developer/CommandLineTools

• Install Homebrew + Cakebrew
• Install via homebrew (from science tree)

  cmake
  suite-sparse

Cygwin

Thanks to the autor of this post for showing the way for matplotlib.

• Install cygwin (basic system) following the usual setup instructions given on cygwin
• Using setup_x86_64.exe, install the following additonal packages:

gcc-g++
wget
curl
openblas
cmake
lapack
libumfpack-devel

python
python-numpy
python-devel 
swig

• If you get tired clicking into the setup gui, download apt-cyg and copy the script to /usr/bin (which is C:\cygwin64\usr\bin from the DOS prompt). Then you can install all the packages using apt-cyg install.
• This should be sufficient to run make and make test. Before that, close the cygwin shell and open it again in order to catch the proper path settings for lapack
• For Python graphics (matplotlib), proceed further:
  • Install the cygwin packages

libfreetype-devel 
python-pyqt4
xorg-server
xinit

• Obtain pip, the python package manager

$ wget https://bootstrap.pypa.io/get-pip.py
$ python get-pip.py

• Install matplotlib via pip:

$ pip install matplotlib

• Start the X server and set the DISPLAY environment variable

$ startxwin
$ export DISPLAY=:0.0

• use qt4agg as python backend:

$ export MPLBACKEND=qt4agg

Installation of numcxx

TU Berlin UNIX Pool

Here, the current version of numcxx, without the examples, is installed in the directory /net/wir The example subdirectories can be copied separately as a whole.

Virtual Machine

The debian-numcxx virtual machine can be downloaded from the course home page. Load this machine into VirtualBox and start it. Vtk, numcxx and vtkfig and codeblocks are installed on this machine. Examples can be run with the numcxx-build utility.

Build on your own system

Go to the numcxx root directory and set up the cmake project:

$ mkdir .build
$ cd .build
$ CXXFLAGS=-std=c++11 cmake ..
$ cmake --build .

This should compile the library and make it findeable by depending projects via an entry im the user CMake project registry found under .cmake in your home directory.

Working with the examples

First, copy an example directory to your working directory

$ cp -r <numcxx>/examples/whatever .
$ cd whatever

Each of the example subdirectories has its own CMakeLists.txt which looks for an installed numcxx during the setup phase.

Plain unix

Set up a build subdirectory and compile the project:

$ mkdir .build
$ cd .build
$ CXXFLAGS=-std=c++11 cmake ..
$ cmake --build .
$ ./<whatever example>

Code::Blocks

Set up a build subdirectory for codeblocks:

$ mkdir .build
$ cd .build
$ CXXFLAGS=-std=c++11 cmake -G"CodeBlocks - Unix Makefiles" ..
$ codeblocks <whatver the project name is>.cbp

In codeblocks, select a target under "Build/Select Target". In the management pane (left), select the corresponding source file. Now you can build and run via the correspondin menu items of codeblocks.

Other IDEs

Many of them should work. Like with code blocks, find out the name of the generator via cmake -G. If you don't find your IDE name there, there might be some chance that your IDE supports cmake directly (as e.g. QTCreator does).