[Openchrome-users] VIA Chrome 9

[Pavel Krejcir]

Jon Nettleton wrote:
> On Wed, 2007-06-13 at 23:21 +0200, Pavel Krejcir wrote:
>   
>> Hello everybody,
>>
>> I've got an MSI board with VIA chipset VT3230, with integrated graphics 
>> card K8M890. After some investigation I've found a 2D driver supported 
>> by OpenChrome, and realized that neither a 3D binary driver nor a 3D 
>> specification has been published by VIA, and it is likely that it never 
>> will be. (Not sure what "Vista Certified" really means, I'm afraid that 
>> it could be decrypted as "The vendor promises that free software will 
>> never be supplied with appropriate information.")
>>
>> I apologize for a bit lengthy  description at this point, I want to be 
>> sure that they who are interested in will properly understand what I am 
>> actually asking for.
>>
>> Well, since I would sometimes like to have a 3D acceleration, a simple 
>> solution might be to buy another graphics card which I could be sure a 
>> driver exists for. On the other hand, this is not a critical problem for 
>> me. I will probably never use the 3D acceleration heavily, sometimes try 
>> to draw some things in Blender, nothing more. Of course, this is still 
>> possible with disabling DRI, but direct rendering would be better. I 
>> believe the card is not as bad as stated on some forums, and for this 
>> purpose should be fully sufficient.
>>     
>
> The 3d engine for the Chrome 9 chipsets is very different than the
> existing Mesa drivers.  The engine is a programmable core and so far we
> have no info on it.
>
>   
>> So I started to investigate the problem a bit. I decided not to ask 
>> anything on any forum prior getting some basic knowledge. After few 
>> weeks I know that the direct rendering for this card is disabled in DRM. 
>> I can locate the point and switch it on. After recompiling, installing 
>> drm.ko and via.ko and reloading the system, the X gets frozen every time 
>> I start the screen saver settings, or run glxinfo. The X is frozen so 
>> much that even "Ctrl-Alt-Backspace" does not help. That's fine, and I 
>> expect it to do this. I have another computer in LAN so I can safely 
>> reboot the affected machine via telnet. I am prepared for real debugging.
>>     
>
> There is no reason that DRI can't start working.  Our problem was that
> we were having getting DMA working on the IGP/AGP bus.  Once we get that
> straightened out we will re-enable DRI.
>
>   
>> Eventually I start to approach the proper topic of this message. I am a 
>> little bit stacked now. I know that parsing of the buffer does not cause 
>> the crash, it is caused by something called next, when the buffer 
>> parsing returns OK. I know that the buffer parsing is called somewhere 
>> from mesa, but I cannot compile mesa. There is some syntax error in a 
>> module probably not used by my computer at all. I should be able to 
>> overcome this and compile mesa in the end, but I am not sure whether it 
>> is even necessary.
>>     
>
> You will definitely need to be able to compile mesa.  The mesa
> openchrome driver handles all the unichrome/chrome9 3d functionality.
>
>   
>> There are at least three components - drm, mesa and openchrome - which 
>> somehow collaborate and I do not fully understand how. My guess is that 
>> I should focus rather on openchrome than on mesa. So what I really need 
>> is if somebody can shed any light on drm-openchrome relationship. What 
>> library is loaded in order to parse openGL instruction into the driver 
>> command buffer? Is it unichrome_dri.so, or should it be a via_dri.so 
>> (which does not exist on my machine)?
>>     
>
> Openchrome handles all the 2d rendering and acceleration and DMA.  DRM
> handles the direct hardware access.  Mesa handles all the 3d
> acceleration and rendering.
>
> Unichrome_dri.so is loaded to handle the openGL, but like I stated
> above, won't work the chrome9 chipset.
>
>   
>> If you manage to read it through here then you probably guess that I am 
>> ready to experiment with my hardware. I want to try anything I can, 
>> however foolish it might appear. So I will appreciate any kind of 
>> information from anybody who is willing to share it.
>>
>>     
> We do want to support this functionality, but are waiting on information
> from VIA.  The other option is to do a full reverse engineering of the
> 3d engine, which is quite in-depth.  Hope this helps.
>
> Jon
>
>   
>> Thanks in advance.
>>
>> Pavel
>>
>> _______________________________________________
>> openchrome-users mailing list
>> openchrome-users at openchrome.org
>> http://wiki.openchrome.org/mailman/listinfo/openchrome-users
>> Main page:
>> http://www.openchrome.org
>> Wiki:
>> http://wiki.openchrome.org
>> User Forum:
>> http://wiki.openchrome.org/tikiwiki/tiki-view_forum.php?forumId=1
>>
>>
>>     
>
>
> _______________________________________________
> openchrome-users mailing list
> openchrome-users at openchrome.org
> http://wiki.openchrome.org/mailman/listinfo/openchrome-users
> Main page:
> http://www.openchrome.org
> Wiki:
> http://wiki.openchrome.org
> User Forum:
> http://wiki.openchrome.org/tikiwiki/tiki-view_forum.php?forumId=1
>
>
>
>   
Jon,

thank you for your quotes. I am a bit more clever now. I have only two 
more questions:

1. Is there any chance/promise that VIA will supply the mesa team with 
the required specification? And if yes, is there any timeframe when it 
should happen?

2. Is there something I could help with meanwhile? - I think this has 
been already answered by Benno later today.

So thanks again,

Pavel