Over the past few months momentum has been building in the community, here is an attempt at providing a round-up of current software projects and I’ll apologise in advance for any omissions!
Dr Beau Webber has been working on adding Epiphany support to the APL to C compiler, aplc. To find out more about the array-oriented language and for details of progress made and code examples, see the APL forum.
Omer Kilic and Edward Tate of Erlang Solutions have been exploring how Erlang can be used together with OpenCL to implement vision systems which make use of the Epiphany accelerator. They introduced an object tracking project back in May and further details will be posted here soon.
There has been a great deal of interest in using Parallella with software-defined radio (SDR), and Tommy Tracy II (tjt7a) has started work on porting GNU Radio signal processing blocks to Epiphany. Once again, watch this space for further details.
Mortdeus has made a start on implementing Go support for Epiphany, and if this is something that you would like to contribute to, the Go forum is the place to get involved.
When Hoernchen posted details of his LLVM Epiphany back-end in May we were seriously impressed, and we were delighted when yabin also started contributing to this effort.
There is much more work to be done before we will have a robust LLVM implementation and if you would like to get involved contact Hoernchen via the LLVM forum.
Message Passing Interface (MPI) is used extensively in high-performance computing (HPC) applications, and a few weeks ago simsciloki started work on an implementation for Epiphany.
The Epiphany SDK provides a functional simulator for a single Epiphany core, but earlier this month Gravis announced that they were starting work on a multi-core emulator.
Over a series of posts censix described how to build R — the language for statistical computing —on Parallella, and to access the Epiphany accelerator from within R using the ROpenCL package.
Becoming a Contributor
If you’d like to contribute to one of the above efforts or have an idea for a new project, post to the relevant forum (or if you prefer you can contact me first via e-mail: andrew@).
If you’re looking for inspiration here are just a few of our priorities:
- Improving our U-boot fork
- A Parallella optimised SETI@home
- Hadoop support for highly energy-efficient big data applications
- A Parallella metadata layer for Poky (for custom Linux distros/images)
- Python/NumPy/SciPy support for the Epiphany accelerator
A Great Start
Considering that only a modest amount of Parallella prototype hardware is out there in the field this is a really great start, and imagine what our software ecosystem is going to look like 3-6 months after 6,300 boards have shipped!