XDC: randr extension for full display state setup (full set of outputs and crtcs in one shot)

Dave Airlie airlied at redhat.com
Mon Oct 5 17:18:30 PDT 2009


On Mon, 2009-10-05 at 20:12 -0400, Alex Deucher wrote:
> On Mon, Oct 5, 2009 at 7:54 PM, Dave Airlie <airlied at redhat.com> wrote:
> > This was in ajax's summary and I just wanted to raise the issues with
> > this before anyone seriously started thinking about it.
> >
> > 1. We should not fail modesetting unless its an extreme situation, we
> > should endeavour to provide the remote client as much information as
> > it requires to pick a set of modes the gpu can actually set. It this
> > is just a workaround for this case then you should fix the randr
> > protocol to allow smarter client situations.
> >
> > 2. multi-seat on a single card - what happens if we give a crtc to each
> > user with outputs partitioned between them? what if a single X server
> > isn't running all the bits of the card, or has them all exposed.
> >
> > I'd be interested in the situations where something like this is
> > actually required, i.e. I'm sure it can allow for smoother mode
> > setting when someone swaps connectors, but thats probably not the
> > common case.
> 
> Arron and I were thinking mainly for mode validation across multiple
> monitors.  You may have two or more huge monitors connected to a
> low-end card that can't drive both at full res due to bandwidth
> limitations.  It can however drive one monitor at full res just fine.
> So how do we deal with this?  Which modes on which monitors do we
> reject or do we keep them all and have one of the monitors flicker or
> blink or not come on due to underflow.
> 

Okay so in this case can we expose this information to the client
so it can make an intelligent decision? what do you do if this is
X's startup modesetting? have it just fail to start?

If the GPU detects two monitors plugged in, how about we limit
the mode list somehow?

I'm just not sure how failing the modeset is going to help no
matter what you do, how is the client expected to know what was
wrong with what he asked for? you need to enable intelligent
client behaviour much more than fail.

Dave.




More information about the xorg-devel mailing list