XBURST SIMD

Maurus Cuelenaere mcuelenaere at gmail.com
Fri Aug 21 07:56:27 EDT 2009


Hi Wolfgang,

Wolfgang Spraul schreef:
>> I think there's no way Ingenic is going to publicly release their 
>> documentation regarding the MXU instruction set..
>>     
>
> Why not? Are you talking to them? :-)
>   
No, I'm not :)

But I do have the MXU instruction set documentation (although a leaked 
one), and the person that gave it to me really stressed on the fact that 
I shouldn't release/leak it to the public (which I'm not going to).

> I have met with the CEO of Ingenic numerous times and we have a good
> relationship. I have also started talking about SIMD with him.
> The thing is that I don't always want to demand new things without
> delivering anything.
> So my strategy is this:
>
> First get them to be more open about the regular Linux BSP.
> They have since, upon our request, released a formerly closed driver,
> mtdblock-jz. I think it's useless and will disappear as part of our
> kernel cleanup, but nonetheless. Then they have released new 2.6.27
> patches (all of this is up on downloads.qi-hardware.com).
> I am hoping they will publicly release the Programmer's Manual for
> all their SoCs. Right now only the datasheets are open, an old version
> of the 4740 PM got leaked, see
> http://www.mediafire.com/?djnnnmnojzd
>   
I know of these and I have them, but there are some missing chapters in 
them (which I got from an other source).

Also that documentation isn't pretty good at some points (like the 
UDC/USB documentation is really vague for some stuff).

I also don't really consider leaked datasheets 'open'.

> However, Ingenic has a new version of the PM (which they also worked on
> upon our request thankfully), but we got it confidentially so I
> don't want to leak it. I hope Ingenic publishes the PM, it would be
> a good next step.
>   
I would be really interested in that :)

> Then my goal is to show them that the Linux community is real by doing
> some work on their Linux BSP. They need help, we are all not perfect.
> Ingenic does not employ the kind of kernel hacker that could get their
> source quality to where you could even try to send patches upstream.
> Luckily it seems some of the Qi activities pick up steam and Mirko Vogt
> and Lars are starting to do kernel hacking. It would be very important
> if others could join, maybe Ignacio Garcia from dingux.com when he's back
> from vacation, or others.
> Please join!
> http://github.com/lindnermarek/openwrt-x-burst/tree/x-burst
>   
I currently don't have the time to do much hacking, but I am interested 
in this project.

I'll be watching the mailing list and try to help out wherever I can 
with my (limited) experience of porting the Jz4740 to Rockbox though.

> If we can show a clean up-leveled (2.6.3x) kernel development to them,
> I can guarantee you they will open up more.
>
> Sure they consider SIMD to be their 'crown jewels', but it's also clear
> when other semiconductors add SIMD extensions (Intel SSE, AMD 3DNow!),
> of course they publicly and professionally document them. I cannot see
> why Ingenic would not want to do this, it's just a matter of doing all
> the work and doing it right.
> They are working on new things, SIMD2, 3D (opengl), GPS, etc.
> In all of this we (Qi Hardware) will want 'open documentation', but we
> can only really make the case if that 'open documentation' leads to
> something, namely software development. Perfectly reasonable to me.
>
> Next step is a great Linux kernel. After that I will go back to SIMD.
> Hope this makes sense, feedback welcome, and most importantly anybody
> who can join kernel hacking, please do so...
> Wolfgang
>
> P.S.: Worst case, if Ingenic remains closed for good in some areas, we
> at Qi Hardware will eventually have to look for alternatives, really open
> chips, maybe even 100% GPL'ed chips like www.opensparc.net. But right now
> Ingenic seems to be responsive, and I feel the ball is on our side to
> get some work done...
>   




More information about the discussion mailing list


interactive