ATI Radeon KMS Leaves Staging Area In 2.6.33

[glisse]

KMS is still not ready in 2.6.33. It's simply too slow and full of graphical glitches (I'm on an RV770). But it's good that it's out of the staging area, as any promising driver should.

You opened a bug for this ? What puzzle me is that i am completely unable to reproduce all reported rendering issue despite the various r6xx/r7xx i have.

[pvtcupcakes]

You opened a bug for this ? What puzzle me is that i am completely unable to reproduce all reported rendering issue despite the various r6xx/r7xx i have.

KDE's Kwin window manager has some glitches, but I don't know if it's Mesa or DRM.

Sometimes a window gets a thick Yellow or Black border around it. If you focus or unfocus the window it goes away, but it'll eventually come back on another window.

[ivanovic]

KDE's Kwin window manager has some glitches, but I don't know if it's Mesa or DRM.

Sometimes a window gets a thick Yellow or Black border around it. If you focus or unfocus the window it goes away, but it'll eventually come back on another window.

Jepp, the same over here. On my rv670 I do especially get "black rectangles" where eg the currently running program is shown or the whole area for switching the current workspace or the clock just shown that black rectangle. Things somehow seem to be related to stuff like transparancy. Clicking once on the desktop tends to fix those things. And when borders around progs are shown clicking on a different prog also fixes this issue.
Sorry, but it is rather hard to provide you with some way to definitely be able to reproduce it. But at least here using kwin (kde 4.3.x atm) I do see those issues. Nothing that stops you from normal working, just some glitches. Those only happen with KMS active and it makes no difference if acceleration method for desktop effects is xrender or opengl. So I think the "fault" is somewhere in the KMS part, not in mesa or libdrm itself, though I am not sure at all.

Currently I am running 2.6.33-rc5, but I had the same observation with KMS in 2.6.32*. Using mesa, libdrm and xf86-video-radon from git-master.

[pingufunkybeat]

Yeah, I have the same problem.

It shows up on and off. In the past, it was a Mesa commit that caused it, then it was reverted and things were OK. But it's there again.

[Obscene_CNN]

I personally think KMS has quite a way to go on catching up with UMS in performance. As I have hacked away on both sets of code to improve performance by any means I just don't see where KMS is going to get it from or where the core bottle neck is. Minor improvements are possible however.

[agd5f]

Note that there are some EXA bugs in xserver 1.7 that can cause these sorts of problems.

[humanoid]

bug in kms radeon
http://bugzilla.kernel.org/show_bug.cgi?id=15166
in 2.6.33-rc6 not fixed
kernel full freeze and system not ping from remote computer

[humanoid]

mesa from git and xf86-video-ati from git

[humanoid]

oops =(
kernel 2.6.33-rc6
system full freeze and not ping from remote computer

http://bugzilla.kernel.org/show_bug.cgi?id=15166

[RealNC]

You opened a bug for this ? What puzzle me is that i am completely unable to reproduce all reported rendering issue despite the various r6xx/r7xx i have.

No, I thought they were known issues. Before I file a bug, are the following a known problem?

http://i48.tinypic.com/12479kw.png
http://i46.tinypic.com/20pa4yc.png

(Black and yellow boxes where a smooth shadow should be.)