From Qi-Hardware
Jump to: navigation, search


Qi Hardware makes innovative projects. Here are a few highlights.

[edit] Copyleft Hardware

  • copyleft hw = hardware built only with freely available knowledge.
  • the first computer built out of Wikipedia

[edit] What makes copyleft hardware superior?

Well you know the long laundry list, question is what sticks

1. no vendor lock-in 2. you can max out software upgradability, drive the actual physical bits to their maximum lifetime 3. dramatically lowers the amount of capital/people needed to improve upon the piece of hw, distributing the innovation burden on more shoulders 4. foundation for software continuity, i.e. you can make investments into or on top of software, and not be worried that your foundation is gone one day 5. will be cheaper than proprietary hardware once the amount of reusable copyleft hw reaches a critical point that makes investing to replace it prohibitively expensive at that point (#5), it should really take off btw. but I think it's quite a way out, especially because proprietary hw can be subsidized with proprietary content or ads, better than copyleft hw

[edit] Ben WPAN, akak SLoWPAN

[edit] First Open Processor

Milkymist1 is the first open source system-on-a-chip. The software that runs on our FPGA is completely GPL'd.

  • We democratize IC design
  • Make it accessible to anybody

IC design is where the innovation in the next 20 years will come from OS design is maxed out application design (whatever model-view-controller stuff or similar you want) languages - bah - we have enough compiled, interpreted, jit, etc. IC design is the forgotten child, huge potential once it is pulled into the software equation

speaking about that - what innovation have you seen in the Linux kernel in the last 10 years?

Its mature, Its polished!

IC design is also the entry door to DSP (digital signal processing), which is the entry door to many applications, RF protocols being one of them also ultrasound, microwaves, radar, etc.

imagine if your smartphone would also be an ultrasound device, the implications for global health would be unimaginable

in order to reach those applications, we need to look beyond what we have in software right now like I said above, beyond 1) OS design 2) application design 3) programming languages those things are all maxed out database design all digital stuff MAXED OUT if you wait 10 years, don't expect innovation to come from there object-oriented databases? come on UML - unified modeling language

Sebastien, the Linus of IC design, made the first steps

you can also say with IC design, we can marry the digital world with the analog world that's innovation in itself, and more once it is used in actual applications

[edit] Rapid Design

  • Over the weekend you can hack up something cool

[edit] Hardware is Much Cheaper than Equal System

As pure hardware, one innovation is that it is cheaper than other systems that can do the same thing especially if you want to control DMX, MIDI as well

[edit] What applications can FPGA and Milkymist be used for?

> What types of applications could you think of for writing verilog code
> for the milkymist soc fpga that could be interesting out side
> ofgraphics?

Hmm, let's see. First of all, the FPGA is useful for all sorts of
"self-referential" work. E.g., to prototype new peripheral
interfaces in the Milkymist product line. We don't make much use
of this at the moment, but that may change.

Candidates for such interfaces could also include unusual things
like those GPS chips.

Second, the FPGA can act as a universal coprocessor. This could
range from all sorts of codecs to crypto, and so on. Since our
system is already FPGA-centric, we may have more flexibility in
how we integrate that coprocessor.

Third, and that's where I see the main interest beyond the VJ
station, the Milkymist project provides a basis for advancing
the state of the art when it comes to openness in FPGA use (from
a developer's point of view).

It's hard to predict what will come out of "opening" FPGAs, but
I think there's a lot of potential. Directions currently
completely unexplored include on-the-fly synthesis of subsystems
(think classifiers for network packets), or instrumentation like
my hack that brings out FPGA-internal signals (and that is
severely limited by the bugs in unsupported Xilinx tools).

What the Milkymist project can contribute there is not only the
opening of the core processes, but also provision of a reference
platform. One problem with Free FPGA work is that a lot is
centered on development boards. These boards have only a very
limited lifetime and aren't nicely integrated boxes. A Milkymist
platform that with a long-term evolution could make a big
difference there. Of course, at some point in time, the VJ
focus would become a burden. But if we could, say, spin a
MilkyNote off Milkymist and NanoNote, that could go a long way
towards that goal.

> Some guys were asking me about writing optimized code for doing big
> data crunching and more, just want to have some far out thoughts on
> what milkymist could be bent to do for some investors/high level
> thinkers to crunch on...

Not so sure about data crunching. The M1 isn't the fastest
machine when it comes to raw processing power. So you may run
into limits there (memory bandwidth, I/O bandwidth, CPU core

Perhaps it's better not to try to sell the M1 as a box that can
do all sorts of marvellous things, but the Milkymist line, with
the M1 as the box that unlocks the secrets that today keep us
from making that box of marvels.

Cheers, Werner

From Wolfgang:

> Perhaps it's better not to try to sell the M1 as a box that can
> do all sorts of marvellous things, but the Milkymist line, with
> the M1 as the box that unlocks the secrets that today keep us
> from making that box of marvels.

Yes, Milkymist and M1 today are about flexibility beyond what the
current computing architectures offer. Not more raw computing
power, but flexibility to take computing into completely
unexplored and un-optimized for use cases. It's hard to narrow it
down more at this point, but fpga and ic design are the enabling


yeah I am wondering myself about that a lot too :-)
Werner, please correct...

1) graphics acceleration
2) digital signal processing, that is the processing of radio protocols,
or radar/sonar/ultrasound
3) more parallelism, like 10 cameras

Let's see what Werner says...
Personal tools