David Dawes wrote:
>
> On Tue, Apr 02, 2002 at 07:49:59PM -0700, Jens Owen wrote:
> >Alan,
> >
> >I've committed a slew of header file changes. Mostly, I've seperated
> >any device dependencies from drm.h and the automatic include of all the
> ><device>_drm.h files. This results in a drm.h which does not have any
> >of the driver specific IOCTLs, and the need to explicitly include the
> ><device>_drm.h files where needed.
> >
> >Also, I've removed xf86drm<Device>.c files and I've got the
> >xf86drmCompat.c support working for the drivers that haven't been
> >converted to the new drmCommand interface.
>
> Hi Jens,
>
> What happens with xc/lib/GL/dri/drm, where the xf86drm<Device>.c files
> used to get symlinked?
The compatability module: xf86drmCompat.c is linked to that directory
for now...so the build still works for the drivers that haven't been
converted to the new drmCommand interface.
I plan on removing this link when all the drivers are converted, as the
compatability module is not needed for backwards binary compatability on
the 3D driver side. The 3D drivers are staticly bound with the libdrm.a
support they need.
> I was wondering if the xf86drm<Device>.c files would just get moved
> to the ddx driver directory, and be modified to use the new interface
> instead of ioctl().
The primary function of the old xf86drm<Device>.c modules is to convert
from a OS independent interface to an OS dependent IOCTL call. Keeping
the drm<Driver><Function> semantics of the old interface is typically
redundant after removing the OS dependent IOCTL call.
A typical flow using the old interface looked like this:
driver calls internal routine for DRM service
internal routine packs up request for drm<Driver><function> call
drm<Driver><function> repacks request for ICTL call
With the new drmCommand interface, the flow looks like this:
driver calls internal routine for DRM service
internal routine packs up request for drmCommand call
We could certainly add the extra layer back in, it would make things
much easier to port for new drivers--but I think the resulting code is
more convoluted and has more overhead.
> I haven't had a chance to checkout your new branch yet, so maybe
> that would answer my questions.
Radeon is converted, and I'm about to commit the i810.
-- /\
Jens Owen / \/\ _
[EMAIL PROTECTED] / \ \ \ Steamboat Springs, Colorado
_______________________________________________
Dri-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/dri-devel
