Intel X.Org Driver Now Handles Better Tear-Free

[phoronix]

Phoronix: Intel X.Org Driver Now Handles Better Tear-Free

Chris Wilson of Intel this morning released an updated xf86-video-intel X.Org driver. With this latest 3.0 pre-release comes support for TearFree on transformed outputs. There's also been other changes...

http://www.phoronix.com/vr.php?view=MTQ5Mjg

[deanjo]

Sadly, the little black box corruption still exists with this. Wish they would finally fix it.

https://bugs.freedesktop.org/show_bug.cgi?id=68410

[Bucic]

Any chance this fixes the tearing issues for which a workaround was to disable culling in Gnome Shell for example?

http://phoronix.com/forums/showthread.php?84536-How-to-contact-MESA-devs-on-some-pesky-performance-bug&p=359687#post359687

[ickle]

Sadly, the little black box corruption still exists with this. Wish they would finally fix it.

https://bugs.freedesktop.org/show_bug.cgi?id=68410

I will have a workaround of some form before 3.0.

Any chance this fixes the tearing issues for which a workaround was to disable culling in Gnome Shell for example?
http://phoronix.com/forums/showthrea...687#post359687

TearFree would prevent all of that - but really this is a bug in the compositor and DRI2 protocol and should be fixed there.

[Bucic]

I will have a workaround of some form before 3.0.



TearFree would prevent all of that - but really this is a bug in the compositor and DRI2 protocol and should be fixed there.

How do you know? Can you post me to any existing bug reports? I'm struggling with tearing for months now!

[Bucic]

And what is this tearfree you speak of? It doesn't seem to be a package or a tool.

[ickle]

How do you know? Can you post me to any existing bug reports? I'm struggling with tearing for months now!

The crux of the issue is the attempt to "optimise" updates of the frontbuffer through the use of DRI2CopyRegion (called through MESA_copy_subbuffer). That operation is not specified as being synchronous wrt vrefresh, so it is allowed to tear. Furthermore as a tearing DRI2SwapBuffer itself calls the backend CopyRegion function, there is no way the backend can detect whether or not it should prevent tearing. It's a glorious snafu.

However, if the compositor just does a plain old SwapBuffer, the update is tear free. (Of course that in itself has negative implications for prime and slaves like udl.)

The root cause here is the inadequacy of the DRI2 protocol for efficient and tear-free compositing.

[Bucic]

I have no idea what you are talking about In Fedora 17 I had no tearing and now I have. I would just like to know against what package I report the bug and how should I title it.

[Shirudo]

And what is this tearfree you speak of? It doesn't seem to be a package or a tool.

It's a configuration option to remove screen tearing. From the Arch forums:

cat /etc/X11/xorg.conf.d/20-intel.conf
Section "Device"
Identifier "Intel Graphics"
Option "SwapbuffersWait" "true"
Option "AccelMethod" "sna"
Option "TearFree" "true"
EndSection

I've used these options myself and they work perfectly, I think I get better performance from switching to sna too.

[sharan]

when will opengl 3.3 and these feature get in xf86-video-ati