[Mesa-dev] [RFC] Change dri state tracker GetBuffersWithFormat format encoding

[Thomas Hellstrom]

On 12/05/2011 11:38 PM, Eric Anholt wrote:
> On Mon, 05 Dec 2011 15:21:57 +0100, Thomas Hellstrom<thellstrom at vmware.com>  wrote:
>    
>> Hi!
>>
>> Currently the DRI state tracker sends the bits per pixel (bpp) value in
>> the format member of a
>> getbufferswithformat request, and assumes it can reinterpret the format
>> of the returned buffer to something else
>> with an identical bpp.
>>
>> That doesn't really work for vmwgfx. We can't reinterpret a surface
>> format (for example from ARGB to XRGB) and
>> I would therefore like to change the dri state tracker to send the depth
>> instead of bits per pixel, so that the X server driver
>> knows exactly what format it is supposed to create:
>>
>> Color buffer:
>> depth 32 : argb32 (bpp 32)
>> depth 24 : xrgb32 (bpp 32)
>> depth 16 : rgb16 (bpp 16)
>>
>> Depth buffer:
>> depth 32 : z32 (bpp 32)
>> depth 24 : x8z24 (bpp 32)
>> depth 16 : z16 (bpp 16)
>>
>> Depth_Stencil buffer:
>> depth 32 : s8z24: (bpp 32)
>>      
> So how do you distinguish based on depth between a8r8g8b8 and r11f g11f
> b10f from GLX_EXT_fbconfig_packed_float?  I would think you want enums
> along the lines of MESA_FORMAT_* instead of trying to guess formats
> based on a bpp/depth.
>    

We can't distinguish that with this proposal.
The dri state tracker's visual currently supports a very limited number 
of formats and
we cover all of these with this change, which is not very intrusive.

However, I agree with you that it would be desirable to be able to 
better support a wider range of formats, and
that leaves two possibilities:
1) Follow your suggestion. All DDX drivers that want to support the dri 
state trackers would need to implement the server side and
we would need a common definition somewhere. Probably in the dri2 
protocol headers.
2) Keep it driver-private and have the dri state tracker call into a 
driver-private configuration function that translates between gallium
formats and dri2 format codes.

Meanwhile I'm posting a patch along the lines of the original proposal 
so that we at least can distinguish between the formats that are 
currently supported by the dri state tracker.

/Thomas