Has anybody looked at XMOS

Wolfgang Spraul wolfgang at sharism.cc
Wed Jan 6 20:48:17 EST 2010


Ron,

> Having said that, it's not clear to me that XMOS would need
> to reveal the deep details of how it is designed. Nanonote uses an
> Ingenic SOC where Qi Hardware lives with a programming model, and
> reasonably decent documentation. However you don't require Ingenic to
> show you every layer of their masks. <grin>

Oh sure I require that, but even though I talk to the CEO regularly, he
dismisses this of course and talks about his upcoming IPO that will make
him rich :-) So be it...
The Ingenic CPUs are part of how we try to launch this business. With
real products that hopefully (in the short term!) produce real value for
normal people. As you know Ingenic XBurst chips are found in tens of
millions of media-oriented consumer electronics, for a number of reasons.
I feel relatively good with these chips right now, our relationship with
Ingenic is rocky but honest.
How long we work with them depends on them, as long as they see themselves
mostly as a manufacturer it can go on for a long time. If they become very
iffy about all the great IP in their chips, then we will kick them out
earlier. Right now better support for XBurst is in the best interest of
free software, as those chips are used in _LOTS_ of devices as you know.

If our goals are too lofty and we are in the copyleft ivory tower for too
long, this project will fail. That doesn't mean that we don't have our
eyes clear on our goal of producing 100% copyleft hardware. When I say
100% I mean 100%. Including production testing software and tools, the
whole EDA suite actually, etc.

Hope this makes sense...

> The interesting question (to me) is must we have every detail at the
> gate level, or can great and interesting hardware use some high level
> chips which we admit we can
> not reproduce.

Yes we do, that's copyleft hardware.
The point is that _ALL_ information needed to produce hardware is copyleft
licensed. So you have physical goods on one side, and freely available and
licensed information on the other. Nothing in between. We'll get there.
Over time I believe this will create the most interesting, highest quality,
most green, and most profitable hardware.

> Copyleft hardware ~could~ be defined to include commercially
> available parts where the programming interface is described in
> sufficient detail that open software drivers
> can control key parameters.

Commercial yes, free should be commercial otherwise it cannot take off.
A good test is whether a chip is available (commercially) from at least 2
financially independent companies. If it is, at least you have a real
standard (say for SDRAM or NAND chips). Even then you would slowly want
to go down because otherwise those standards with proprietary information
underneath will grow on you.
This test comes out false for the Ingenic XBurst chips, and it seems also
for XMOS.

> But if Qi Hardware pursues an totally purist philosophy, it excludes
> some VERY interesting hardware.

Yes, totally purist please. 100% orange juice :-)
It's just by far the most fun to work in a group of people that doesn't believe
in the value of hiding or monopolizing information.
Unfortunately the FSF and Richard Stallman seem somewhat unwilling to wake up
to the challenge today's super complex yet proprietary hardware poses. rms
still insists on his 'circuits vs software' idea last time we spoke. If they
don't wake up to the reality of how ICs are designed and made nowadays, free
software will suffer until they do.

Like David from tuxbrain told me a few days ago: Information that you hide
is like shit, but if you put it out there something will grow on it ;-)
Wolfgang

P.S.: Sorry for the long ideology mails, I appreciate Ron's input and maybe
once in a while we need to repeat the current state of our copyleft thinking.
It's also evolving... Feedback welcome...

On Wed, Jan 06, 2010 at 05:32:46PM -0800, Ron K. Jeffries wrote:
> I noticed the XMOS is used in the new Amiga is all.
> Reconfigurable processors are intellectually interesting.
> 
> I did not in any way intend to suggest Qi Hardware not continue to
> pursue its goals.
> 
> Having said that, it's not clear to me that XMOS would need
> to reveal the deep details of how it is designed. Nanonote uses an
> Ingenic SOC where Qi Hardware lives with a programming model, and
> reasonably decent documentation. However you don't require Ingenic to
> show you every layer of their masks. <grin>
> 
> The interesting question (to me) is must we have every detail at the
> gate level, or can great and interesting hardware use some high level
> chips which we admit we can
> not reproduce.
> 
> Copyleft hardware ~could~ be defined to include commercially
> available parts where the programming interface is described in
> sufficient detail that open software drivers
> can control key parameters.
> 
> I don't have a dog in this fight. But if Qi Hardware pursues an
> totally purist philosophy, it excludes some VERY interesting hardware.
> 
> But you know that. ;)
> 
> be well,
> ---
> Ron K. Jeffries
> 
> 
> 
> 
> 
> 
> 
> 
> 
> On Wed, Jan 6, 2010 at 16:24, Wolfgang Spraul <wolfgang at sharism.cc> wrote:
> > Ron,
> >
> >> Have you looked at programmable processors from
> >> http://www.xmos.com/
> >
> > http://www.xmos.com/legal/terms
> > "you agree that you will not copy, reproduce, alter, modify, create derivative
> > works, or publicly display any content from this site, other than..."
> >
> > Qi Hardware is a COPYLEFT HARDWARE project.
> > I will not spend time to add more proprietary IP to our products. Not sure
> > anybody else from the core contributors will, but I have some doubts :-)
> >
> > Our project is about removing proprietary IP and build the best copyleft hardware
> > devices in the world.
> > If you think we should look at XMOS, can you ask them the following:
> >
> > *) license all source codes of their chips under GPL (Verilog/VHDL)
> > *) publish documentation about their chips under CC-BY-SA or GFDL
> > *) use only free tools to develop their chips (Fedora Electronic Lab)
> > *) publish documentation about production parameters and steps, testing
> > software and setup, under GPL or CC-BY-SA licenses so that other companies
> > can produce improved versions of their chips
> >
> > We will probably go down the road of doing one or two FPGA boards (Xilinx) this
> > year as a first step towards our own copyleft ASICs. So we see the proprietary
> > Xilinx FPGA technology as a development platform. To me personally, as long as
> > the longer term goal is clear and we are moving towards copyleft ASICs I can
> > live with that.
> >
> > What do you like about XMOS? What exactly would you want to use those chips
> > for? Why not Xilinx FPGAs?
> >
> > Wolfgang
> >
> > On Wed, Jan 06, 2010 at 10:42:44AM -0800, Ron K. Jeffries wrote:
> >> Carlos (and others)
> >>
> >> Have you looked at programmable processors from
> >> http://www.xmos.com/
> >>
> >> at the right price, XMOS would be VERY interesting.
> >> Lots of compute power, lots (!) of i.o
> >>
> >> ---
> >> Ron K. Jeffries
> >>
> >> _______________________________________________
> >> Qi Developer Mailing List
> >> Mail to list (members only): developer at lists.qi-hardware.com
> >> Subscribe or Unsubscribe: http://lists.qi-hardware.com/cgi-bin/mailman/listinfo/developer
> >
> > _______________________________________________
> > Qi Developer Mailing List
> > Mail to list (members only): developer at lists.qi-hardware.com
> > Subscribe or Unsubscribe: http://lists.qi-hardware.com/cgi-bin/mailman/listinfo/developer
> >
> 
> _______________________________________________
> Qi Developer Mailing List
> Mail to list (members only): developer at lists.qi-hardware.com
> Subscribe or Unsubscribe: http://lists.qi-hardware.com/cgi-bin/mailman/listinfo/developer




More information about the discussion mailing list


interactive