Installation

The MinKit package allows to minimize probability density functions (PDFs) using a python API, offering the opportunity to use iminuit or scipy minimizers. The PDFs are compiled at runtime, allowing any user do define custom PDFs.

Installation with PIP

This package is available on PyPi, so to install it simply type

pip install minkit

To use the latest development version, clone the repository and install with pip:

git clone https://github.com/mramospe/minkit.git
pip install minkit

Remember that you can also install the package in-place, something very useful for developers, by calling

git clone https://github.com/mramospe/minkit.git
pip install -e minkit

This package uses the GNU scientific library (GSL), which needs to be accessible in order to compile the source code generated at runtime. In addition, the C++ standard used is C++11 or greater. Be sure to have the necessary environment variables set before running any script. In Ubuntu, this can be done executing:

sudo apt-get install libgsl-dev
export CFLAGS="$CFLAGS -fPIC -std=c++11 -I/usr/include"
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/lib/x86_64-linux-gnu"

Installation with CONDA

This package is also available in conda. However, the dependencies are located on a different channel to that of this package. The dependencies live in the conda-forge channel. In order to properly handle the different channels, it is recommended that you run

conda config --add channels conda-forge

which is equivalent to

conda config --prepend channels conda-forge

Alternatively, you can also execute

conda config --append channels conda-forge

to give the new channel the lowest priority. Afterwards you can simply run

conda install -c mramospe minkit

in order to install MinKit.

Notes for GPU compatibility

This package is capable to work in CPU and GPU backends, and has been designed to work in both CUDA and OpenCL. The GPU operations are done using the reikna package. In order to make MinKit run in GPUs, it becomes necessary to have installed reikna, and pycuda or pyopencl depending if we have CUDA or OpenCL installed in our system. The dependencies are not automatically handled by pip or conda, so you will need to run

pip install reikna pycuda pyopencl

or

conda install -c conda-forge reikna pycuda pyopencl

in order to install them.