Installation¶

The easiest way to install magpy is using the conda repositories. The only requirement is that you have anaconda or miniconda installed on your system. Alternatively, you can build the C++ and python code from source. You might want to do this if you only want the C++ library or if you would like to access the Intel accelerated code.

Conda¶

Note packages only exist for Linux and Mac OSX.

Go to https://conda.io/miniconda.html and download the Miniconda package manager

Create a new conda environment and install magpy into it:

$ conda create -n <env_name> python=3
$ source activate <env_name>
$ conda install -c owlas magpy

Launch python and import magpy
```
$ python
$ >>> import magpy
```

From source¶

The instructions below will guide you through the process of building the code from source on Linux.

Clone the magpy code

$ git clone https://github.com/owlas/magpy

You’ll need a C++11 compatible compiler (g++>=4.9 recommended)
You will also the LAPACK and BLAS libraries. On Debian systems these can be obtained through the apt repositories
```
$ apt install liblapacke-dev
```
Build the magpy C++ library with with your compiler as <CXX>
```
$ cd magpy
$ make CXX=<CXX> libmoma.so
```
You can build and run the tests from the same makefile
```
$ cd magpy
$ make CXX=<CXX> run-tests
```
To build the python interface you’ll need to obtain all the python dependencies in the environment.yml file. Install miniconda (see Conda, above)
```
$ conda create env -f environment.yml
$ source activate magpy
```
Once you have all of the dependencies you can install magpy
```
$ CXX=<CXX> pip install .
$ python
$ >>> import magpy
```

From source (Intel compilers)¶

Magpy has been optimised for Intel architectures and you can take advantage of this by taking a few extra steps:

Clone the magpy code

$ git clone https://github.com/owlas/magpy

Ensure you have the Intel compilers in your path (icc and icpc)

Tell magpy where to find your MKL files

$ export MKLROOT=/path/to/mkl/install/directory

You will also the LAPACK and BLAS libraries. On Debian systems these can be obtained through the apt repositories
```
$ apt install liblapacke-dev
```
Build the magpy C++ library with the intel compilers. The correct build flags should be taken care of for you
```
$ cd magpy
$ make CXX=icpc libmoma.so
```
You can build and run the tests from the same makefile
```
$ cd magpy
$ make CXX=icpc run-tests
```
To build the python interface you’ll need to obtain all the python dependencies in the environment.yml file. Install miniconda (see Conda, above)
```
$ conda create env -f environment.yml
$ source activate magpy
```

Once you have all of the dependencies you can install magpy

$ cd magpy
$ CC=icc CXX=icpc pip install .
$ python
$ >>> import magpy