Parallella: An Open Source Hardware Project 18


We are proud to say that we have now published all the hardware sources for the first version of the Parallella board on Github, making the Parallella a proper open source hardware project. The first beta version of the board will go out to early backers soon! We have also updated the detailed specification for the final Parallella board. This version will be manufactured in volume, with over 6,300 boards being shipped to Parallella Kickstarter backers later this summer. The new design files are in process and will be be published openly as soon as they are ready.

The meaning of “open design”

The definition of “open” in the realm of hardware and software has a long and contentious history. Instead of giving you my own definition, I would rather refer to the following pieces from Stallman, Creative Commons, and freedomdefined.org [1-3]. Our goal for the Parallella hardware project is to set a new standard for open collaboration on a global scale. Here is a summary of the steps we have taken so far in terms of “open” and “free” hardware to date.

  • Licensing:
    • GPL/LGPL V3 for all major software components and Parallella FPGA HDL sources [4]
    • MIT license for small code examples [5]
    • Creative Commons Share alike license for Parallella schematics and documents [6]

Why Publish Now?

Initially our plan was to only release the design files after the Parallella board had reached production status.¬† This way we would have the flexibility to make changes up to the last minute and we could “look professional” in the process.¬† Holding early and formal design reviews with a select few critical colleagues is a necessity. The thought of sharing a potentially inadequate design with the whole world just seemed too painful! We are now quite happy with the state of the Parallella design and we could really use a few thousand sets of eyes to review the design and make it perfect! We have a small window to make changes to the design before the design files have to ship off to the factory. Today we begin the process of converting the Parallella into a true community based project. We look forward to your feedback¬† and criticism. Any constructive feedback will serve to make the the project better.

Sincerely,

Andreas Olofsson

————————

[1] http://www.gnu.org/philosophy/free-sw.html

[2] http://creativecommons.org/about

[3] http://freedomdefined.org/OSHW

[4] http://www.gnu.org/licenses/gpl.html

[5] http://opensource.org/licenses/MIT

[6] http://creativecommons.org/licenses/by-sa/3.0/legalcode


Leave a Reply

18 thoughts on “Parallella: An Open Source Hardware Project

  • Reply
    Clark Hartsock
    This project is a success already. You are doing it right, and leading by example. I am incredibly pleased to be a contributor!
  • Reply
    JRP
    A "new standard for open collaboration" is a half-truth at best. Parallella isn't open until the Epiphany HDL source code is released.
  • Reply
    pandastorm
    I hate to say it, but JRP's right. I'm not saying you should just give out your hard-earned IP. Lord knows the open source model doesn't work for chip design - very few competent people, pretty closed ecosystem, huge entry costs etc. but you can't really call it an open source hardware project though if the main feature is proprietary.
  • Reply
    Andrew Back
    Hi @JRP, @pandastorm, Arduino, BeagleBoard and BeagleBone are all widely recognised as being open source hardware and you don't get the processor HDL sources with any of those platforms. Remember, the Parallella computer is an open source hardware project and no such claims are being made about the Epiphany chip (for which an awful lot more documentation has been made freely available than for many other chips!) We could debate what is and what isn't open (source) hardware, but this has been done many times before elsewhere, and while it would be an interesting exercise it will have no bearing on what has become commonly accepted and defined [1]. Regards, Andrew [1] E.g. http://freedomdefined.org/OSHW & http://openhardware.org/
  • Reply
    Andreas Olofsson Post author
    Andrew: Well said! JRP: I wrote that our "Our GOAL for the Parallella hardware project is to set a new standard for open collaboration on a global scale." Surely a "goal" can't be a half truth? I agree that we still have more work to do in opening up the platform.
  • Reply
    shodruk
    JRP, pandastorm, Even if you use OpenCore's open source IP, FPGA is also full of proprietary technology. If you want a completely open source hardware, you need enormous money. If you have the money, please make it for us. :-)
  • Reply
    Morgaine
    To qualify as open hardware, the board design and the components of a board just need to be fully and openly documented so that new versions of the board can be developed, and of course the software needs to be open source. There is no requirement that the individual components of an open hardware board be open hardware at the gate level, only that their interfacing details and operation be fully specified so that you can link hardware to them and program them. After all, you're going to be buying the devices from their manufacturer, not create them yourself in a foundry. Complaints would only be justified if the Epiphany device lacked interfacing details or had to be programmed or accessed with the help of binary blobs. That would definitely not be open source, but that's not the situation here, AFAIK. The Zynq device has some issues about openness in respect of programming its FPGA since that requires proprietary Xilinx software, but complaints about that should be laid at Xilinx's door, not Adapteva's.
  • Reply
    ali8
    What about Long Term Support? How knows if we buy 10 boards now, will the company be supporting them after a couple of years or no? I think there should be a clear log term support plan, by stating until when will the board be still supported, by bug fixes, new updates/upgrades, daughter boards, SDKs, etc.
  • Reply
    Andrew Back
    @ali8 This is an open source project and, just like Arduino, BeagleBoard and BeagleBone etc, it is community supported. If you want an OMAP eval/dev kit with commercial support you can buy one from TI and it will probably cost you 10-20x times what a BeagleBone does. Similarly, if you want commercial support from Adapteva, I imagine they would be happy to discuss your project requirements with you. Although I expect they are going to have their hands full until much later in the year, and given they have circa 6,300 boards to ship, now might not be the best time to ask. The great thing is that with projects such as this all the toolchain and driver sources etc. are out there, with modifications to upstream projects pushed to them where possible. So, in cases where a project sponsor/lead ceases development, the community has the means to continue support. Cheers, Andrew
    • Reply
      ali8
      @Andrew, Thanks, but I did not mean commercial support. I simply mean that I will still see the product tree five years from now. I don't want to come back 5 years later and find that parallela is a history and my projects need to be modified all over for a new board/kit. Without that, you even cannot talk about open source project, etc. I hope I am clear...
      • Reply
        Andrew Back
        @ali8 So the question is about roadmap and long term availability, rather than support per se. I think it's safe to assume that Adapteva are in this for the long haul. The intention is very much to develop a sustainable ecosystem, and Adapteva's own long term success will be closely linked to the project objectives of "closing the knowledge gap in parallel programming", so that we can tackle some of the tougher challenges associated with parallel computing. This won't happen overnight. Regards, Andrew
  • Reply
    Morgaine
    One of the most important benefits of open source software and open hardware is risk reduction or elimination. You are no longer at the mercy of a corporation declaring a product End Of Life and leaving you in the lurch with a wasted investment in something that is no longer supported. The community or other companies are always able to provide support or continue development. That may be what ali8 meant above about long-term support. Parallella is open hardware, but there is only a single supplier for its Epiphany device so there is still an element of risk remaining despite being open hardware. While Xilinx is unlikely to disappear, the Zynq on Parallella could conceptually be replaced by any other ARM SoC plus a small FPGA or CPLD, but if Adapteva disappears or is purchased by a corporation that has no interest in openness or in continuity then no alternative exists, and there could be a problem for the Parallella community. While people will vary in their degree of long-term worry (or not at all), the issue of single Epiphany supplier does nevertheless exist so I think it was a reasonable question. It would certainly affect any company that decides to clone the open hardware and sell a similar or modified product, since they would be completely dependent on Adapteva for Epiphany devices.
  • Reply
    Remy Dyer
    This project is as close to the ultimate goal of open-sourced hardware as is presently considered possible. Sadly, all chips are "closed source". And until the business model of "closed source + IP circus" is eliminated as a viable business practise, all chips will probably continue to be closed source in design. At present this is a terribly dangerous thing to do, because it is exactly counter to good science and engineering practise. As an engineer, I can only live in hope that this massive social experiment doesn't force future generations to eventually consider the technology to be impenetrable deep dark magic. The "bus factor" for how to actually build chips is already far too low. It's not unprecedented either, already, we (electrical engineers) have largely forgotten how to design beam-forming electrodes for beam power tubes. The practical knowledge, the gotchas and tricks, are already closed-to-lost. Nowadays we have this "religion of the black box", where we have faith that the engineers who designed it really did do perfect work. This is not always true, and chip manufacturers will sometimes prefer to "save face" by omitting disclosure of certain errata. Already it is too late to both keep IP law, and open source chips. The complexity of the cross-licensing agreements make this impossible. The only way it will ever happen, is if the entire world - or enough of it to matter - abandon such legal "anti-features". (Prisoner's dilemma). Reality itself is generally good enough to thwart our aims most of the time, we don't need to maintain the additional arbitrary barriers that are "IP" law. We already know that cooperation is a more viable survival strategy than competition, else science would have no benefit. So I commend these guys - they've put their arse on the line. No doubt there will be copy-cat ripoffs. But those ripoffs will never really compete because they will lack the detailed understanding acquired during development, necessary to assist the community that will form about this powerful new device. This thing is far more than a mere appliance, it's a unique new tool.