Free software friendly graphics GPU? (Kickstarter fundraiser in progress)
I am a member!
http://www.kickstarter.com/projects/725991125/open-source-graphics-processor-gpu
Has anybody come across this yet?
Sent an email inquiring about the goals and clarification on this being a 100% free software solution or if non-free pieces will exist.
If we get an answer I'll update here.
Sure, I'm a regular Phoronix reader :). I would like to know what is the real benefit of these open hardware designs (except that independent companies can manufacture them, which means more competitiveness, which in turn also mean better pricing), since I only care about, if the particular hardware is working within a fully-free environment, or not.
I think that if Intel will be also cooperative on the Coreboot side, by releasing the required microcode sources necessary for proper hardware initialization, we will probably have everything we need (beyond hardcore gamers, seeking raw performance). The free nouveau, and radeon drivers are still better than any of these open designs, but who knows what will happen in the future. Maybe they will also pass the FSF hardware certification program, and we will have a FSF approved GPU :)).
I am a member!
It doesn't sound like it is free software friendly... but it may be? I guess someone will have to look for the VGA BIOS code. He implies it may be available.
Here is what I got back from the project lead:
--
Christopher,
Our GPU is 100% compatible with the Number Nine I128 driver.
This can be found on X.org.
We also have an example implementation for OpenGL.
Our VGA BIOS is a binary which also supports VESA 3.0.
The VGA is 100% IBM VGA compatible, so it may be possible to find
BIOS source code somewhere.
In any case all of our software and hardware will be free if we are
successful.
Regards,
Jim MacLeod
President
Silicon Spectrum, Inc.
The Bochs / Qemu VGA BIOS is free (LGPL).
http://bochs.sourceforge.net/cgi-bin/lxr/source/bios/VGABIOS-lgpl-README
I checked the license of the i128 driver in the Trisquel repo and it is also free.
Duly backed, thanks Chris.
"Our VGA BIOS is a binary ... it may be possible to find BIOS source code somewhere."
From the first part, they only provide a binary? The second part seems as if they're blowing you off. It's "somewhere" - go look for it, whatever - we don't care - we're not distributing it - we're only going to provide the binary.
If the Bochs / Qemu VGA BIOS source code can be used why don't they and why don't they distribute that? If the Bochs / Qemu VGA BIOS is free and can be used then that's great but that their response is as if they don't care is what concerns me.
There's also the matter that most FPGA still have non-free software needed. In a way we'd be trading the non-free software currently used in the GPU world with the non-free software used in FPGA development and programming. It seems the non-free stuff just morphs & moves. That's not necessarily a reflection on these folk though; just something to think on.
So not everybody has to read a long post I'll do this in British 'top
line' format, the crux first. One question here at the top for
everybody. Do you want something more than the Lemote Yeelong as a
[now barely] usable free software friendly computer platform? I
thought so, well as you know libre doesn't mean gratis.
Here's the detail:
The Kickstarter's product is Verilog HDL _source_ licensed under
LGPLv3, so it's free and not limited to FPGAs; verilog can be used
with many VLSI tools. The 'Respects Your Freedom hardware
certification requirements' [1] say:
> However, there is an exception for secondary embedded
> processors. The exception applies to software delivered inside
> auxiliary and low-level processors and FPGAs, within which software
> installation is not intended after the user obtains the
> product. This can include, for instance, microcode inside a
> processor, firmware built into an I/O device, or the gate pattern of
> an FPGA. The software in such secondary processors does not count as
> product software.
So it's even possible to have a proprietary FPGA in the board design
and the computer still be RYF compatible.
In RMS's GNU 30th speech he said words to the effect 'support the
free hardware community, theirs might be the only hardware we can
trust in the future.' This OTOH is an open source hardware project.
My view is like the open source software community they're useful for
the freely licensed products they make. After all we're using a
derivative of Linus Torvald's kernel and you know what he's like.
The kickstarter is potentially an important libre component of a
modern computer the free hardware community can use.
AIUI the open source hardware definition is about zero licensing cost
not liberty. So the fools would just go and buy a 'cheap'
proprietary license on an extant i128 VGA BIOS without a bye or leave.
I've backed this not because ultimately I want to buy a PCIe card
with this GPU to stick in my x86 PC. I've backed it for the
'creation of available free platforms for Linux (sic)' this undated
free hardware page [1] sees as 'unstoppable whatever happens.' That
the free software community has to port and write software for it
because of the open source hardware designers' lack of ethics seems a
minor inconvenience when compared with the fact that the free
hardware community currently has no choice but to use proprietary EDA
tools against their ethical principles in pursuit of this goal. If
you read the page [1] you will see there is a team working on a libre
FPGA to rescue them, and therefore us, from this.
It might seem a lot of money in terms of having it to spend in your
personal bank account, but, the maximum $1M stretch target for a
modern GPU definition works out as approximately 13 elapsed months at
the standard UK 'time actually working' ratio for just 3 (sic three)
ordinary software contract staff in the UK. The company I worked for
at the time was spending more per year on some modest (teens of
people) proprietary software teams and their overheads when RMS
changed the world and published the four freedoms just over thirty
years ago. In terms of chip design today $1M is utterly peanuts.
Even if this company had free software principles I cannot see how
they would fit the VGA BIOS and 3D GPU driver development into such a
_tiny_ budget given the task at hand. To me it seems natural that
the drivers for free hardware be written by the free software
community. The software is, after all, our half of the free computer
equation.
Bear with me for a bit. Last year's 'free and open source hardware
like the Arduino' Kickstarter the Parallella which some of us backed
IMO turned out a complete success in free hardware terms. We have a
novel parallel computer architecture fully documented, SDK and
libraries including a libre OpenCL implementation, complete HDL to
interface the parallel CPU cores to a dual core 1GHz ARM, complete
circuit board design etc all libre.
Unhappily those of you who would rather we not develop free hardware
using the only tools there are, are going to be upset because part of
what makes it IMO a complete free hardware success is it's FPGA which
is on the CPU chip is soft loadable. Thus it will never be RYF
because of the proprietary FPGA tool chain. But non-free FPGA tool
chains are according to all the sources I've found _all_ there is.
At some point in time there has to be a free hardware GNU/Linux
capable platform which is coupled to an FPGA so that free peripheral
IP and the corresponding free software including drivers can be
developed. The Parallella is this within the limits of what can be
done with free software and free hardware at present.
AIUI even Coreboot contains non-free microcode or your x86 CPU won't
work, and AFAICT most of us are still using the proprietary BIOS in
our computers from lack of a coreboot image for it. We regard that
as an acceptable compromise because there is no RYF computer at
present. I've chosen a different compromise, a free hardware
computer where the compromise is the non-free FPGA toolchain used by
the free hardware community. The whole of the source is libre. Well
it will be when there's a fully free ARM distro until then it needs
Debian porting (because Adapteva ship it with Ubuntu) and before that
testing with a deblobbed kernel.
The Parallella Kickstarter was again $1M to turn it into a reality,
some of us backed it. Its slipped schedule but is we're told is
shipping this month. They're taking pre-orders from non-backers for
it now, $99.
The above wasn't entirely a disgression, at the 600K stretch goal the
open source hardware GPU Kickstarter will build project files for the
Xilinx ZYNQ series. That's the ARM/FPGA combo in the Parallella.
I've not checked up to know for sure the IP will fit in the
Parallella with necessary other bits of libre IP. Adapteva intend to
build different models with different gate count FPGAs if there is
demand. So the possibility of an 1GB dual core ARM SoC with a
performance libre GPU coupled with a cutting edge 16 core parallel
processor is there. Parallel computer technologies are a potential
future escape from current CPU design limitations when we hit the
limits of silicon physics. So that too is an important libre
hardware and software bet for the future.
I fully accept the veracity of the RYF definition, hardware that
fully respects our freedoms has those properties. In essence all I'm
suggesting wrt to the Parallella is that we listen to our allies in
the free hardware community when they talk about the realities of
hardware design. Remembering in addition that RMS has said free
software is a long haul cause which implies we should keep our eyes
on the goal at the horizon when we're weighing up a choice between
two compromises.
The above is only the result of me researching and writing up my
inferences for something a little more than a day. So by the way of
these things it needs checking. The opinions here on the Parallella
are strictly my own.
So please Chris and others back this. Even without the Parallella
this is an investment against the day we are no longer beholden to
proprietary hardware companies.
"AIUI even Coreboot contains non-free microcode or your x86 CPU won't
work"
Not really. You just don't get bug fixes. It's not that the CPU refuses to work at all.
"We regard that as an acceptable compromise"
Perhaps some do. I don't - proprietary software is never a solution - and have removed such from my coreboot system.
Yes, it would be nice to see an FPGA that can be used without any proprietary software anywhere. I'm not sure anyone's working on that, especially if they accept that "that's all there is" and find it to be an acceptable compromise.
Some CPUs are too buggy to run much more than the microcode update,
devices with them should be avoided. Some need workarounds in kernels
and toolchains, which aren't always well-documented.
I believe there was a project replacing the nonfree program for one FPGA
model, I cannot find it now.
> Perhaps some do. I don't - proprietary software is never a solution
> -
> and have removed such from my coreboot system.
For all cases I can think of other than this one in current times you
are quite right. However, in this specific case the free hardware
community are bootstrapping free hardware replacements for
proprietary hardware. This is IMO entirely analogous to RMS
bootstrapping the GNU project using proprietary development software.
However, I quite understand if YMMV.
"proprietary software is never a solution"
Reflecting on this some more as I went for some groceries. There's I
was paying for things from a computer controlled refrigerator
(non-free software) using a debit card in a reader (non-free
software) through a POS terminal (non-free software) using my bank
account (non-free software) in an air conditioned shop (non-free
software).
So, until the free software movement is further down the line,
'never' is mere cognitive dissonance.
More like - not on my devices that have GNU/Linux built for them
(Replicant devices use non-free code in the modem/DSP side of the
hardware) unless I have absolutely no other choice and am absolutely
certain I'm not indulging in what Satre called 'bad faith' when I
make that choice. And obviously I buy free software capable devices
when there is a choice.
Or to put it more simply - I always seek to maximise my software
freedom. Which, I think, is what will be your reality too.
> So, until the free software movement is further down the line,
> 'never' is mere cognitive dissonance.
They may be using non-free software but that doesn't invalidate what I said - It's still not a solution. Their customers merely interacting with it doesn't make their customers responsible for it being there and they don't need to feel bad for having interacted with it. Perhaps sorry for the shop owner for the decisions they've made.
>> So, until the free software movement is further down the line,
>> 'never' is mere cognitive dissonance.
>
> They may be using non-free software but that doesn't invalidate
> what I
> said - It's still not a solution. Their customers merely interacting
> with it doesn't make their customers responsible for it being there
> and they don't need to feel bad for having interacted with it.
> Perhaps
> sorry for the shop owner for the decisions they've made.
I absolutely agree, I'd missed your intended meaning of the word
solution. To rephrase for clarity, software that does not give the
user their four freedoms is never the right answer because it is
always abusive to users (your meaning). However, because I have no
choice but to be paid electronically, indirectly using a bank's
proprietary software is the only answer I'm allowed (the meaning I
read).
I've written to Frank Bruno (the project lead) and got a
clarification of the position with respect to the Parallella. The 2D
part (which can be used separately) will fit in the 7010 equipped
Parallellas, it might just fit in the 7020 equipped ones, but if it
does it'll be tight.
Also they've moved creating a generic interface for the code to the
200K base target to make it more useful to people.
The result of these two is if the kickstarter makes its first funding
target the Parallella will gain 2D accelerated graphics.
I am a member!
I wonder what the following means?
"Software drivers are a challenge, and we will work on providing some level of drivers, with the hopes that the community takes them up and pushes them to new levels and provides problem reports to us."
"Some level of drivers"?
That seems to be the case way too often.
I like the idea, but I'm not clear on who it is for. Gamers? Regular users? "Free aficionados? The direction is not clear, at least to me.
Hopefully if you receive an answer to your email, it will make more sense.
I am a member!
I think it depends on how much support they can drum up. If it's only 200,000 then it'll be 2d then there is no advantage for most over what we currently have (probably). There may be special use cases though for a niche audience. If it is a million they think they can get something competitive with commercial offerings from INTEL/NVIDIA/AMD. Competitive might be too strong of a statement. They use the word "modern" though which to me implies at least early generation voodoo cards (some of the first 3d accelerated cards). They do state this is targeted at the desktop too if I'm not mistaken.
What makes this project different than past projects is they have core of it done already and the expertise to follow through to a complete 3d accelerated product if support can be drummed up.
Between what lenny2010 found out about the VGA BIOS situation and the response from Jim MacLeod it sounds like we have a winner as far as free software goes.
I'm prepared to make a financial contribution to this effort.
I won't do so immediately. If anybody thinks there is a reason to object on free software grounds speak up. Otherwise I'll contribute something before the fund raiser is over.