Skip to content

Latest commit

 

History

History
101 lines (78 loc) · 2.69 KB

INSTALL.md

File metadata and controls

101 lines (78 loc) · 2.69 KB
Raw

Installing Pyranda

Prerequisites

Before installing pyranda you'll need to install numpy and mpi4py. Numpy's f2py utility is used to build the python-fortran interface and we try to validate that your choice of compiler will work with the installed mpi4py's version of mpi.

installing mpi4py

mpi4py should be installed with the --no-cache-dir option to avoid using an existing build with a cached compiler.

# if your mpi*s are in your path
pip install mpi4py --no-cache-dir
# otherwise you can specify an environment variable
env MPICC=/path/to/your/mpi pip install mpi4py --no-cache-dir

installing numpy

numpy shouldn't have any special build steps, just install as normal:

pip install numpy

Installing with pip

[optional] Using a virtualenv

...> virtualenv [-p python] my_venv
...> source my_venv/bin/activate
(my_venv) ...>

You can also verify that you're in your venv by checking your $PATH:

...> echo $PATH
/path/to/your/env/root/my_venv/bin:...

install pyranda

pip install . [--user]

Installing without pip

[python setup.py build [extra_build_args]]
python setup.py install

Legacy Instructions - Manual Install

This process should work on any system and will allow for an arbitrary compiler to be used for the fortran and for the mpi4py.

Step 1: Ensure python and numpy

Python-

Though other versions of python may very well work, we recommend and support python 2.7, 3.5, and 3.6 for pyranda.

numpy-

As long as numpy is working with your version of python above, there will be no compability issues. This can be installed in a number of ways. http://www.numpy.org

Step 2: Custom install of mpi4py

This python package provides MPI bindings to python and may or may not exists on your system and python path.

Install mpi4py (this should work on most systems with a mpi compiler installed)

wget https://bitbucket.org/mpi4py/mpi4py/downloads/mpi4py-3.0.0.tar.gz
tar xvzf mpi4py-3.0.0.tar.gz
cd mpi4py*
python setup.py build --mpicc=/where/you/have/mpicc
python setup.py install --prefix=install_location_mpi4py

** Add install_location_mpi4py/*/site_packages to PYTHONPATH **

Step 3: Pyranda build/install

A fortran compiler compatible with the mpicc used in mpi4py is used by default.
2003 and above standards enforced and MPI libraries is required.

Install pyranda

git clone https://github.com/LLNL/pyranda.git
cd pyranda
python setup.py build
python setup.py install --prefix=install_location_pyranda

** Add install_location_pyranda/*/site_packages to PYTHONPATH **

Step 4: Run tests to check install

Trying navigating to pyranda/examples and running

python advection.py