modular: Changes to 'master'

[Luc Verhaegen]

On Tue, Oct 21, 2008 at 05:45:59PM +0100, Daniel Stone wrote:
> 
> So what you're saying is that it's the exact same as -radeon for newer
> hardware.  For older hardware, the details of how it achieves the same
> end result differ.

The details are still _very_ different, even for newer hardware.
 
> The last time we had this discussion, you hadn't actually looked at
> RandR 1.2, but had already decided it was crap.  Have you looked at it
> this time? Have you any plans to create anything better than RandR 1.2,
> or is this abstraction layer exactly that -- another abstraction layer?

We're still trying to work out how to deal with mapping DCE 3.2 hardware 
to RandR and its limited view of a modesetting layout. I'm sure that 
once the hardware arrives, and we work around enough, we will be able to 
make it somewhat work, but it will be far from pretty.

> > > It now has EXA, DRI and Xv code copy and pasted
> > > from Radeon.
> > 
> > Try again.
> 
> Please correct me if I'm wrong?
>
> > > There is the CS (command submission) infrastructure, so if you
> > > desperately want 3D support without a DRM, radeonhd is the market leader.
> >
> > You clearly haven't been watching this code at all.
>
> Please correct me if I'm wrong?

R5xx 2D acceleration has its origins with -radeon driver code, but it 
has been reworked to a high degree and now is based on the CS 
infrastructure. Only 3 files are active copies; radeon_3d.c, 
radeon_exa_render.c and radeon_textured_videofuncs.c; all contain very 
R5xx specific code.

> > So what stops it from being shipped as well?
> > 
> > Nothing. Just you.
>
>

Luc Verhaegen.
SUSE X Driver Developer.