Experimental options

[Alex Deucher]

On Mon, 13 Dec 2004 20:23:27 -0500 (EST), Vladimir Dergachev
<volodya at mindspring.com> wrote:
> 
> 
> 
> 
> On Mon, 13 Dec 2004, Alex Deucher wrote:
> 
> > On Mon, 13 Dec 2004 18:53:36 -0500 (EST), Vladimir Dergachev
> > <volodya at mindspring.com> wrote:
> >>
> >>
> >> On Mon, 13 Dec 2004, Ian Romanick wrote:
> >>
> >>> So, there's three separate issues here.
> >>>
> >>> 1. Should the DDX use the DRM to assist with 2D acceleration?
> >>
> >> They already do, at least all radeons. Or maybe I misunderstood your
> >> question ?
> >>
> >
> > well, the drm is only used if 3d is enabled.  but if you don't have
> > enough ram for 3d or you don't want to use the dri, the driver falls
> > back to MMIO.  It'd be nice to use the drm for 2d all the time
> > assuming the drm is available for your platform.
> >
> 
> I admit I don't know the dri code that well, but it seemed to me that
> depth buffer (for instance..) was allocated dynamically in a function
> that allocates context, right ?

no, the back and depth buffers are allocated statically when the server starts.

> 
> On the other hand the 3d is enabled at the start of the Xserver, so it
> should remain enabled whether there are any DRI clients or not.
> 
> Or am I missing something ?

yes, but if 3d does not initalize for any reason, the DDX defaults
back to using MMIO for 2d accel rather than the CP.

Alex

> 
>                       best
> 
>                          Vladimir Dergachev
>