[ltp] Re: Radeon and 1680x1050

Alex Deucher linux-thinkpad@linux-thinkpad.org
Mon, 11 Jun 2007 08:12:52 -0500


On 6/11/07, Rolf Leggewie <no2spam@nospam.arcornews.de> wrote:
> Alex Deucher wrote:
> >> It is permanent.  The external screen keeps the higher color setting
> >> where the internal screen looses some color detail.
>
> > I suspect one the CLUT is not being restore properly for crtc1 so the
> > palette is wrong.  Can you file a bug at https://bugs.freedesktop.org?
>
> Sure. http://bugs.freedesktop.org/show_bug.cgi?id=11230

Thanks.  I'll take a look at it when I get a chance.

>
> > Sure it does.  dis-similar head sizes have been supported since the
> > beginning.  you can either have the heads scroll or set the
> > non-rectangular option and they will stay at fixed offsets.
>
> Yes, the scroll thing I knew about.  But that was really not what I
> wanted.  The non-rectangular option and its function are new to me.
> I'll try it out.  So are there any advantages of Xinerama over MergedFB?
>  I can't see any.

well, there are for certain things, but for what most people want,
mergedfb is the preferred method for enabling dualhead (until you try
randr 1.2 that is ;) ).  "Regular"/xinerama/zaphod multi-head mode has
the following advantages/features compared to mergedfb:
- with xinerama turned off you have two independent X screens that can
be a different color depths, etc.
- you can adjust the CLUTs independently
- since you have two separate framebuffers (one for each head) stride
and coordinate limitations are less of an issue since you re-base the
engine to the offset of each buffer.

mergedfb has advantages over zaphod mode:
- DRI works on both heads
- on the fly dualhead/clone mode switching

>
> > So you'd like bottom rather than top alignment?  That can be achieved
> > with the offset option I mentioned before.  However, with that
> > enabled, you will get the current behavior at the top.
>
>
> Exactly, I'd like not to have it on either end.  But it is only a minor
> nuisance.  One that I can certainly live with.

unfortunately there's currently no way around that unless you want to
mess with the cursor warp code.

Alex