Building on The Opportunity at Hand

Parallella is much more than simply a hardware project, and software that is able to harness the power of massively parallel systems is going to be crucial to achieving our goal of making parallel computing accessible to everyone. This includes everything from languages and frameworks that make such systems easier to program, through libraries and toolkits that provide the building blocks for powerful applications, to demonstrators and end-user applications that underline the opportunity.

Ecosystem Foundations

The Epiphany SDK and COPRTHR OpenCL implementation together provide the foundations of our software ecosystem: the compiler and debugger infrastructure, C library, and framework for creating heterogeneous applications that execute across ARM and Epiphany cores.

Only a limited number of prototype boards exist at the time of writing, but we have extensive hardware documentation, source code for the SDK that includes libraries, loader and Linux device driver, and a functional simulator for a single Epiphany core.

We are now in a position to start thinking in more detail about what the missing pieces of our software ecosystem might look like, and work has already been started in a number of key areas.

Work Under Way

Right now members of the community are either developing, or exploring how we might develop, support for:

  • Erlang
  • Python
  • Go
  • Forth
  • An Epiphany operating system
  • GNU Radio

There could easily be efforts that we’ve missed, it’s impossible to know what everyone in the community is working on and some prefer to not share details until certain progress has been made. But suffice to say it’s a really great start!

A Call to Action

If you’re working on something and in a position to do so I’d encourage you to share details via the forums, as this will help to further build momentum in the community and could lead to opportunities for collaboration. Similarly, if you have an idea for something that you’d like to work on but perhaps need some guidance, the forums are great place to find that support.

The Kickstarter surveys helped us identify technologies which we need to make a priority, and those which as far as we know are not currently been worked on include:

  • Computer vision, e.g. OpenCV
  • Hadoop
  • BOINC (as used by SETI@home)
  • OpenMP
  • MPI

If you have experience with any of the above, we need your help!

5 Comments

  • capnrob97 says:

    As far as ‘call to action’ items, I think the BOINC one can generate a lot of buzz.

    I know einstein@home source is available, perhaps someone can start working on porting it to epiphany?

    • Andrew Back says:

      There does seem to be quite a lot of interest for using Parallella with BOINC projects.

      I suspect work may need to be done on BOINC first, before a project that uses it. E.g. to add a Parallella platform. Then a project that provides source, such as einstein@home or SETI@home, would seem like an obvious first candidate.

      Work could absolutely start on this now, and if anyone would like to rise to the challenge we’ll give them all the support we can!

      • capnrob97 says:

        BOINC is already running on the Raspberry PI, so it shouldn’t be too difficult getting it up and running on this platform. Also, Asteroids@home has already ported their science code to run on Raspberry Pi boinc, so I definitely think the interest would be there from some of the projects to support parallella.

        • Andrew Back says:

          Yes, but this would just be BOINC/Asteroids@home running on a generic ARM platform and not making use of the Epiphany accelerator.

          The BOINC work I’d imagined was creating a new platform called Parallella or Epiphany within BOINC, so that when a client connects it knows that there is Epiphany hardware that can be used.

          There would then be work within the individual projects to ensure that they had a version of their application which supported this new platform and knew how to make use of the Epiphany hardware.

          • capnrob97 says:

            I see.

            Hopefully someone runs with this as I see this as a great way to generate some excitement out there for this platform. Too bad folding@home isn’t a BOINC project.

Leave a Reply