[PATCH 3/3] DRI2: defer swap if relevant CRTC is in DMPS off state

[Michel Dänzer]

On Die, 2012-12-18 at 21:25 -0500, Ilija Hadzic wrote: 
> On Tue, Dec 18, 2012 at 1:07 PM, Michel Dänzer <michel at daenzer.net> wrote:
> >
> >      1. Mimic the behaviour with a running CRTC as closely as possible
> >      2. Don't bother, just always wait for a fixed amount of time, say
> >         some fraction of a second
> >
> > I fail to see the point of a flawed compromise.
> 
> Wasn't my first patch series (the one where I always waited
> nominal_vblank_period) implementing option 2?

It's what I'd call a flawed compromise, as it derived the time to wait
from the CRTC's refresh rate, but the result was nothing like the
behaviour with a running CRTC (at least not for anything more complex
than glxgears).


> If constant wait time is acceptable, I can go back to the first patch
> series and simplify it even further (e.g., forget nominal_vblank_period
> and just wait for some fixed hard-coded time). That would address my
> original objective (i.e. avoid dissipating extra power when we turn off
> the screen) and would be very simple.
> 
> If that is not acceptable, and you insist on option 1, then I agree,
> my second patch series needs more work.

In the original thread about this problem, you asked what the right
thing to do would be. I gave my opinion on that, basically option 1
above. I still think that would be ideal, especially if done the way you
outlined, resulting in consistent behaviour between the major drivers.
But if you don't want to go the long way, I clarified above what I think
is the alternative that makes sense. That's all.


-- 
Earthling Michel Dänzer           |                   http://www.amd.com
Libre software enthusiast         |          Debian, X and DRI developer