Catalyst 8.3: Corruption around cursor when using compiz

[dposselt]

Under the last two versions of the fglrx driver, I have had cursor corruption when using compiz (same with beryl, compiz fusion, etc...). The corruption looks like three horizontal lines stretching from the cursor to the right to a distance of (approximately) 60 pixels or so. I am running fedora 7 64-bit with compiz fusion 0.5.2. I tried a fresh install of compiz fusion 0.6.0--no change. Under fedora core 6 running Beryl 0.21 with fglrx 8.40 I get no corruption... Any ideas? Is there something I should be setting in xorg.conf?

Edit: The corruption only appears at the native screen resolutoin of 1680x1050--when I use amdcccle to switch to a lower resolution (any resolution seems to work), there is no corruption around the mouse.

[tibidi]

Hi,

I got the same effect with the same configuration (PC and Soft). Do you have a solution for this ?

Regards

Under the last two versions of the fglrx driver, I have had cursor corruption when using compiz (same with beryl, compiz fusion, etc...). The corruption looks like three horizontal lines stretching from the cursor to the right to a distance of (approximately) 60 pixels or so. I am running fedora 7 64-bit with compiz fusion 0.5.2. I tried a fresh install of compiz fusion 0.6.0--no change. Under fedora core 6 running Beryl 0.21 with fglrx 8.40 I get no corruption... Any ideas? Is there something I should be setting in xorg.conf?

Edit: The corruption only appears at the native screen resolutoin of 1680x1050--when I use amdcccle to switch to a lower resolution (any resolution seems to work), there is no corruption around the mouse.

[dposselt]

I have not yet been able to find a solution for this--it appears to be a problem with the native 1680x1050 resolution. When I connect an external monitor (native 1920x1200 resolution), the corruption around the mouse is not there. When I reduce the screen resolution on the external to 1680x1050 (via amdcccle), the problem still does not appear. I have also noticed that the corruption will go away when the mouse is moved into a window, then reappears when I click inside the window or on the desktop...

I am thinking that this may be a remnant of the problem with screens with native resolution that is not a multiple of 64...? I am hoping that one of the Phoronix members or AMD folks will have suggestions...

Edit: a number of sources have suggested setting Option "XAANoOffscreenPixmaps" "true" in xorg.conf. Doing this did not change anything for me...

[tibidi]

Hi,

Did you got a solution for this ? I tried version 8.4 and I alway have the mouse corruption.

Regards

[bridgman]

I don't think this has anything to do with the "multiple of 64 pixel" issue. If you move the cursor a few pixels (<<64) do the lines move with the cursor or do they stay "pinned" to the screen and jump ahead periodically ?

[shishir]

i don't know what the reason is, but if you were to do "fglrxinfo", you will see direct rendering is not happening. The driver is using mesa 3d libs, instead of ati provided libs. If you run 2-D desktop, i.e., no Xgl, you will see no problems at all. "fglrxinfo" shows ATI as the openGL provider. This is not seen with 8.40.4 driver, but with all the drivers since then, 8-2, 8-3, 8-4.
I have exactly the same configuration as you have: T60p, FireGL v5250, openSUSE 10.3. I am back to using 8.40.4. I dont feel like going to Xorg, as I find compiz useful as well. It is not just desktop bling

[tibidi]

The 3 lines "are attached" to the mouse move even if the move is <<64 pixels.

[dposselt]

The 3 lines "are attached" to the mouse move even if the move is <<64 pixels.

Using catalyst 8.4, I can confirm that fglrx is the rendering agent in this case--I can run 3D applications (I do not game on Linux--these are workstation applications) just fine and fglrxinfo shows ATI as providing the direct rendering:

display: :0.0 screen: 0
OpenGL vendor string: ATI Technologies Inc.
OpenGL renderer string: ATI MOBILITY FireGL V5250
OpenGL version string: 2.1.7415 FireGL Release

Testing glxgears (and fgl_glxgears), I get virtually identical results whether or not I'm running compiz (3000 FPS, glxgears and 700 FPS for fgl_glxgears). Of course, these numbers drop if I start moving the cube, but that is to be expected...

The cursor corruption bug is not a showstopper for me, as I can (finally) suspend/resume on my laptop with the fglrx drivers and run 3D apps, but it is nevertheless pretty annoying. As I had mentioned in a previous post, this only shows up at the native 1680x1050 screen resolution when using compiz or beryl. If I attach an external monitor (1920x1200), there is no problem.

This makes me think that it is still a remnant of the old multiple-of-64-pixel screen width 3D bug, though now my 3D apps look just fine and it is only the mouse cursor under compiz/beryl that is affected.

One last note: when I use metacity as my window manager (as opposed to compiz or beryl), the cursor looks just fine. I have found, though, that an accelerated desktop greatly enhances my productivity...

[shishir]

Hmm...thats weird, I wonder why I don't get same output for fglrxinfo, I am running
xorg-x11-server-7.2-143.11
xgl-git_070104-84.1,

could these be a reason?

[edit]: BTW, I have a T60 as well, with x1400 Mobility Radeon card. I get the same output from fglrxinfo.

[dposselt]

Hmm...thats weird, I wonder why I don't get same output for fglrxinfo, I am running
xorg-x11-server-7.2-143.11
xgl-git_070104-84.1,

could these be a reason?

[edit]: BTW, I have a T60 as well, with x1400 Mobility Radeon card. I get the same output from fglrxinfo.

Hmm... I typically use AIGLX for compiz/beryl, but I do have Xgl installed on my system (xorg-x11-server-Xgl-0-0.7.20070102git.kwizart.fc7). I tried logging into Gnome with Xgl, and also got the Mesa output from fglrxinfo... Strange. Have you been able to use AIGLX instead of Xgl? AIGLX has been working well for me for the last few driver releases...

In /etc/X11/xorg.conf:

Section "ServerFlags"
Option "AIGLX" "on"
EndSection

...

Section "Extensions"
Option "Composite" "enable"
EndSection