Hardware overlay on G965

[Maxim Levitsky]

On Tuesday, 26 February 2008 14:33:41 Thomas Fritzsche wrote:
> Hi Keith and Maxim,
> 
> thanks so much for the update.
> Meanwhile I patched the latest stable ubuntu (7.10) src package (intel
> 2.1.1) and everything seems to work nicely. :-)
> Attached is the "quick & dirty" proof of concept patch.
> 
> I changed:
> * Added: +#define HW_HAS_OVERLAY(pI810) (IS_I965G(pI810) || IS_I965GM(pI810))
> (as I guess X3100 is supposed to have a i965GM?)
> * I disabled the the texture video completely.
> * minimal adjustment to version 2.1.1
> 
> So overlay adapter is used as default and all video application work
> nicely with default settings in compiz.I have this nice little blue
> frame when moving a window, but at least it works.
> 
> Only thing I notice is, that "top" shows me about 20% CPU utilization
> playing a DVD. Is this normal for overlay playback?
> 
> Attached output of "xvinfo".
> 
> Cheers,
> Thomas
> 
> 
> On Tue, Feb 26, 2008 at 3:25 PM, Keith Packard <keithp at keithp.com> wrote:
> >
> >  On Tue, 2008-02-26 at 15:12 +0900, Thomas Fritzsche wrote:
> >
> >  > Also I wonder you just deleted:
> >  > -    /* No overlay scaler on the 965. */
> >  >
> >  > Do you have a clue why this comment was in the original driver?
> >
> >  We thought that the final G965 wouldn't have the scaler, and so we wrote
> >  the textured video code and never tried the scaler code on production
> >  hardware.
> >
> >  --
> >  keith.packard at intel.com
> >
> 
Hi,

My patch is against -git, but it is very trivial, so it should apply on other branches too.

Due to the fact that G965 and GM965 have overlay, and G35 probably have it too,
I guess that I need ether to remove the checks for overlay at all, since all Intel graphical
seems to have overlay, or change this:

+#define HW_HAS_OVERLAY(pI810) (1)

so if overlay gets removed in future versions of those chips only this macro can be changed.

Btw, this macro:
+#define OVERLAY_NOPHYSICAL(pI810) (IS_G33CLASS(pI810) || IS_I965G(pI810))

will need to be adjusted too, since future versions are likely to inherit the behavior of G965.


Best regards,
	Maxim Levitsky

PS:
my next goal is to understand why xrandr 1.0 compat code locks up the chip when used,
randr 1.2 works perfectly (although I have only VGA output on the system.
Recent fix for overlay made it even work in when screen is rotated.