System RequirementsΒΆ

These instructions assume that you intend to install OpenMDAO for normal use – that you will not be hacking on the OpenMDAO infrastructure or standard library, or developing any plugins that require compilation. If you do intend to perform some kind of OpenMDAO development, you should refer to the installation instructions in the Developer Guide because the developer environment has some additional system requirements on certain platforms.

OpenMDAO requires that the following software be installed at the system level on all platforms:

  • Python (2.7.x)
    • In general, OpenMDAO supports the specified versions of Python found at python.org. Alternative Python builds, such as Enthought, ActiveState Python, or Python(x,y) are not officially supported at this time, though some may work with OpenMDAO.
    • OpenMDAO does not support Python 3.x or any Python versions below 2.7. Having the incorrect Python installation in the PATH and/or not having the OpenMDAO-approved Python version in the PATH will only lead to problems.
  • NumPy (version 1.6 or higher)
  • SciPy (version .11 or higher)
  • Matplotlib
  • For the GUI: Chrome Web Browser (version 20 or higher). Note: Things should work in recent versions of Firefox, but we’re not currently testing with that browser. Things will not work properly in Internet Explorer.

Platform-Specific Requirements

We recommend that you read the entire section relating to your desired platform before installing any software.

Linux:

Our current OpenMDAO distribution for Linux is a source distribution, so to use it, you must have compilers (C and Fortran) on your system.

  • On Fedora, the names of the needed RPM packages are:
    • python-devel
    • numpy
    • scipy
    • python-matplotlib
    • gcc-gfortran
  • On Ubuntu, the .deb package names are:
    • python-dev
    • python-numpy
    • python-scipy
    • python-matplotlib
    • gfortran

Note

For our Ubuntu Pangolin users, be aware that Numpy and SciPy versions that you attempt to sudo apt-get install may report back as being the latest available version but will still be beneath OpenMDAO’s threshold. The system is showing you the latest version available on the servers that support Pangolin, but the version will not meet the minimum requirements needed by OpenMDAO. According to the Unbuntu Software Center: “Canonical does not provide updates for Scientific tools for Python. Some updates may be provided by the Ubuntu community.”

Fortunately, you can circumvent this problem in a few easy steps:

  1. Download the packages from SourceForge.net

  2. sudo python setup.py install from each package. At this point, you’re close, but Numpy and SciPy aren’t importable by Python yet, because the Python installer restricts default access rights. To get the right permissions on the packages, continue to step three.

  3. Change permissions inside the distribution with these commands.

    cd /usr/local/lib/python2.7/dist-packages
    sudo find scipy -type d -exec chmod o=rx {} \;
    sudo find scipy -type f -exec chmod o=r {} \;
    sudo find numpy -type d -exec chmod o=rx {} \;
    sudo find numpy -type f -exec chmod o=r {} \;
    
Please note: Some users have reported that this method of installation may somehow uninstall Matplotlib. In the event that occurs, you may have to install Matplotlib in the same fashion as the packages above.

Windows:

Note

Numpy and SciPy have one-click installers. You should use those unless you have a very good reason not to. For Python 2.7.x you want a SciPy version such as scipy-0.11.0-win32-superpack-python2.7.exe. The version number might be different than the one here, but make sure you get something with superpack and .exe in the name.

If you go to the Numpy link above, you might see: “Looking for the latest version? Download numpy-1.7.0.zip.” You DON’T want to get that one. It’s a source distribution, not the installer. The same goes for SciPy. Click through to the latest version and get the .exe file.

Not required, but highly recommended:

This will improve distributed simulation startup time significantly by allowing public/private key pairs to be stored securely.

Mac OS X:

  • Python – Install a new version (2.7.x) from python.org because the built-in version has a distutils bug that will cause some of the OpenMDAO tests to fail.
  • NumPy
  • SciPy
  • Matplotlib

Note

We provide pre-compiled binaries for all Intel OS X Macs, for Snow Leopard or higher. However, if you you’re running an older version of OS X or a PowerPC architecture, then you’ll need to set up some compilers on your system to get OpenMDAO to install. This should be a rare situation, but if you happen to run into it, go to our developer docs for the details about which compilers you need. Get the compilers set up the way we suggest here, but then come back and follow the release installation instructions from the next section of these docs.

OpenMDAO Home

Previous topic

Getting Started

Next topic

Installation

This Page