Re: DRI integration

Michel Dänzer wrote:
> On Fri, 2003-11-14 at 00:05, Keith Whitwell wrote:
> 
>>Lionel Ulmer wrote:
>>
>>>On Thu, Nov 13, 2003 at 03:39:33PM -0700, Brian Paul wrote:
>>>
>>>
>>>>That said, there's nothing to prevent the development of a GLX 
>>>>extension that might expose the framebuffer or off-screen image 
>>>>buffers to clients via mmap'ing or some other technique.  Again, this 
>>>>may not be feasible on some hardware.
>>>
>>>
>>>A bit off topic, but well, said extension (if possible) would greatly help
>>>us poor Wine developpers to handle the 'nice' possibilities of Direct3D
>>>(where, basically, the 3D framebuffer is just another surface which the
>>>application can write to directly). And, even worse, the same is the case
>>>for the ZBuffer (which completely kills performance when emulating it using
>>>GL).
>>
>>I wonder how this works.  The radeon, for instance, doesn't have a linear 
>>layout of the z buffer, so writing to it will not give expected results 
>>(whatever those might be).
> 
> 
> Linear access could probably be achieved with a surface translation
> though.

I'm not sure what they are - is this a hardware facility I've missed?

Anyway, tiling is just the start, things like Z compression and just plain 
caching mean that buffers look less & less like regular surfaces.  I'm not 
sure that this D3D 'feature' will persist for very long...

Keith