Sandy Bridge issue: 3d clipped to window offset (0,0)

Sandy Bridge issue: 3d clipped to window offset (0,0)

[Lindsay Roberts]

Hi,

I'm having some interesting behaviour with Sandy Bridge (Chip
SANDYBRIDGE_M_GT2) and the 2010Q4 package.

3D draws, but the only part of the drawn content I see is the
intersection of the actual current window content rect and the
rectangle from (0, 0) to the width and height of the window at
creation time. Any changes to the position or size of a GL window do
not change this clipped boundary, although the actual content is
offset and/or scaled correctly (but still clipped).

Booting an Ironlake i5 under the exact same software (ie. the same
drive) has no problems. I've been trying to grep for any places the
gen6 path diverges where this could be caused to no effect. Using a
simple window manager, either my own super simple embedded one or twm.

Attempted configs (All other parts as per 2010Q4):

Kernel:
* v2.6.37 tag mainline (has intermittent GPU hangs)
* ickle/drm-intel/drm-intel-next latest
[34da1327c3814781925396fa10c42f596588ff76] (hang-free)

Mesa:
* v7.10 tag mainline
* master latest [469ea695bbbc984ebec26b2413ab70d450a283ca]

Am I doing something absurd with my build/config that could cause
this? In other respects the driver seems to be performing really well.

- Lindsay Roberts

Re: Sandy Bridge issue: 3d clipped to window offset (0, 0)

[Zou, Nanhai]

Hi, 
	We can not see this issue here.
	Could you submit a bug at https://bugs.freedesktop.org/
Thanks
Zou Nan hai

>>-----Original Message-----
>>From: intel-gfx-bounces+nanhai.zou=intel.com@lists.freedesktop.org
>>[mailto:intel-gfx-bounces+nanhai.zou=intel.com@lists.freedesktop.org] On
>>Behalf Of Lindsay Roberts
>>Sent: 2011年1月11日 15:13
>>To: intel-gfx@lists.freedesktop.org
>>Subject: [Intel-gfx] Sandy Bridge issue: 3d clipped to window offset (0,0)
>>
>>Hi,
>>
>>I'm having some interesting behaviour with Sandy Bridge (Chip
>>SANDYBRIDGE_M_GT2) and the 2010Q4 package.
>>
>>3D draws, but the only part of the drawn content I see is the
>>intersection of the actual current window content rect and the
>>rectangle from (0, 0) to the width and height of the window at
>>creation time. Any changes to the position or size of a GL window do
>>not change this clipped boundary, although the actual content is
>>offset and/or scaled correctly (but still clipped).
>>
>>Booting an Ironlake i5 under the exact same software (ie. the same
>>drive) has no problems. I've been trying to grep for any places the
>>gen6 path diverges where this could be caused to no effect. Using a
>>simple window manager, either my own super simple embedded one or twm.
>>
>>Attempted configs (All other parts as per 2010Q4):
>>
>>Kernel:
>>* v2.6.37 tag mainline (has intermittent GPU hangs)
>>* ickle/drm-intel/drm-intel-next latest
>>[34da1327c3814781925396fa10c42f596588ff76] (hang-free)
>>
>>Mesa:
>>* v7.10 tag mainline
>>* master latest [469ea695bbbc984ebec26b2413ab70d450a283ca]
>>
>>Am I doing something absurd with my build/config that could cause
>>this? In other respects the driver seems to be performing really well.
>>
>>- Lindsay Roberts
>>_______________________________________________
>>Intel-gfx mailing list
>>Intel-gfx@lists.freedesktop.org
>>http://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: Sandy Bridge issue: 3d clipped to window offset (0, 0)

[Chris Wilson]

On Tue, 11 Jan 2011 18:12:54 +1100, Lindsay Roberts <lindsay.roberts.os@gmail.com> wrote:
> Hi,
> 
> I'm having some interesting behaviour with Sandy Bridge (Chip
> SANDYBRIDGE_M_GT2) and the 2010Q4 package.
> 
> 3D draws, but the only part of the drawn content I see is the
> intersection of the actual current window content rect and the
> rectangle from (0, 0) to the width and height of the window at
> creation time. Any changes to the position or size of a GL window do
> not change this clipped boundary, although the actual content is
> offset and/or scaled correctly (but still clipped).

That sounds reminiscent of a bug the shadow+dri code has. Check that it is
not falling back to using software 2D and is able to use the BLT engine.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre

Re: Sandy Bridge issue: 3d clipped to window offset (0, 0)

[Lindsay Roberts]

On Tue, Jan 11, 2011 at 9:48 PM, Chris Wilson <chris@chris-wilson.co.uk> wrote:
>> 3D draws, but the only part of the drawn content I see is the
>> intersection of the actual current window content rect and the
>> rectangle from (0, 0) to the width and height of the window at
>> creation time. Any changes to the position or size of a GL window do
>> not change this clipped boundary, although the actual content is
>> offset and/or scaled correctly (but still clipped).
>
> That sounds reminiscent of a bug the shadow+dri code has. Check that it is
> not falling back to using software 2D and is able to use the BLT engine.

Thanks Chris, I can confirm this was the issue. Once I disabled the
(pci.revision < 7) check in the X driver for BLT the issue was fixed.

Chris, Nanhai, do you want me to file a bug against the shadow code
path, or is this enough a known issue?

- Lindsay

Re: Sandy Bridge issue: 3d clipped to window offset (0, 0)

[Chris Wilson]

On Wed, 12 Jan 2011 12:06:14 +1100, Lindsay Roberts <lindsay.roberts.os@gmail.com> wrote:
> Chris, Nanhai, do you want me to file a bug against the shadow code
> path, or is this enough a known issue?

I know about it. I was left scratching my head over it when hooking up
dri+shadow and I get the occasional complaint from unfortunate gen2 users
forced to use shadow.

But I don't have a bug specifically for this.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre