radeon 8500 mergedfb flicker

[Alex Deucher]

On Fri, 03 Dec 2004 01:12:19 -0500, Michel Dänzer <michel at daenzer.net> wrote:
> On Thu, 2004-12-02 at 14:25 -0500, Alex Deucher wrote:
> 
> 
> > On Thu, 02 Dec 2004 12:34:54 -0500, Michel Dänzer <michel at daenzer.net> wrote:
> > > On Wed, 2004-12-01 at 21:12 -0500, Alex Deucher wrote:
> > > > On Thu, 2 Dec 2004 11:05:19 +1100, Dave Airlie <airlied at gmail.com> wrote:
> > > > >
> > > > > Also when I run a 3D app both screen revert to clone mode - wierd....
> > > >
> > > > turn off pageflipping.  it doesn't really work with dualhead.
> > >
> > > I've been scratching my head over the MergedFB AdjustFrame functions but
> > > don't see how both CRTC offsets can end up being the same in
> > > radeon_cp_dispatch_flip(). Maybe the AdjustFrame wrapping is wrong wrt
> > > the generic DRI wrapper? Just a guess...
> >
> > I know I went through the same thing when I wrote mergedfb initially.
> > perhaps something to do with vblanks?
> 
> What do you mean? The flips aren't vsync'd.

yeah, I forgot about that.

> 
> > what's weird is that it used to work ok if crtc1 was at a lower offset
> > than crtc2.  apparently that stopped working after the dri->xorg merge.
> 
> Maybe at the end of RADEONMergePointerMoved(), you should call the
> wrapped PointerMoved function or at least pScrn->AdjustFrame() instead
> of RADEONDoAdjustFrame() directly?

worth a try.  I wanna say I tried that once before and it didn't help,
but it's been a while and may be worth revisiting.

Alex

> 
> --
> 
> 
> Earthling Michel Dänzer      |     Debian (powerpc), X and DRI developer
> Libre software enthusiast    |   http://svcs.affero.net/rm.php?r=daenzer
>