Catalyst 9.2 Released, Still Fails To Deliver XvBA

[bridgman]

FWIW, Xv performs colour space conversion (from YCbCr to RGB) as well as scaling. Might not seem like much but it does offload a couple of gigaflops/sec of work, which is a big deal for most CPUs (and is a big deal for *all* CPUs if the code isn't making full use of SIMD extensions and multiple cores).

[rahman.duran]

FWIW, Xv performs colour space conversion (from YCbCr to RGB) as well as scaling. Might not seem like much but it does offload a couple of gigaflops/sec of work, which is a big deal for most CPUs (and is a big deal for *all* CPUs if the code isn't making full use of SIMD extensions and multiple cores).

Hmm, thanks bridgman, I didn't know that.

[rahman.duran]

here are the xorg.0.log (http://pastebin.com/m57555410) and xorg.0.log.old (http://pastebin.com/m1da4c20b) (I don't know if the logfile is wiped clean after a reboot, but I imagine it is).

here id the xorg.conf I used (I left out a big block of comments, but that shouldn't matter):


Section "Device"
Identifier "Configured Video Device"
Driver "radeonhd"
Option "AccelMethod" "exa" # default shadowfb
Option "DRI" "on"
EndSection



I'll reboot, reproduce the error, and then, before I reboot again, I'll pastebin those logs.

EDIT: I rebooted and looked at the logs. It seems that at the time radeonHD is supposed to be loaded, it thinks that fglrx is already loaded. It is true I have fglrx installed, but I always thought the loading process was done by xorg.conf.
So I'll uninstall fglrx and try again.

you need to wipe out fglrx.

[theviper]

I had to go back to 9.1 because the "black borders" workaround (aticonfig --set-dispattrib=tmds2i,positionX:0 etc.) for 1920x1080 stopped working. Xrandr now says "DFP2" instead of previous "default", and trying to set tmds2i dispattribs with aticonfig ends in error. Doing --query functions I figured out aticonfig now would recognize "dfp" output, but nevertheless I cannot set display attributes for any display name I can think of.

Using HD 3450, HDMI-over-DVI and having the famous underscan (and following hang if you don't reset to amdpcsdb defaults every time you logout) problem. Setting display attributes at every login works with 9.1.
Quite a mess.

If anyone knows how to set the parameters in 9.2, let us know.



Same here ... with 9.2 I was (after a few fglrx versions) able to set the refresh rate without crashing the system.

But the underscan is unbearable ... any1 know how to turn it off?

[MU_Engineer]

So what kind of CPU do you have? Becouse it is nothing to do with XV, there is no acceleration for hd content on amd/ati side neither on fglrx nor foss drivers. It is completely CPU related to play hd videos. So I can't play 1080p videos on my x2 4000+ CPU without frame drops. So if you don't have a fast enough CPU you can't play any hd content both on fglrx and radeon/radeonhd drivers.

I have an X2 4200+ and can play MPEG-2 1080i/1080p video just fine using xv without framedropping (my GPU is a Radeon HD 3850 using fglrx 8.12.) I wonder what the issue is in your system. Where I get into trouble is playing 1080p H.264/MPEG-4 stuff.

[JeanPaul145]

you need to wipe out fglrx.

Did that, installed both the drm driver and the xserver driver, activated radeonhd with the xorg.conf posted earlier...
The performance was abysmal.:D / :(
It would give some empty screen on top of the normal desktop if I ever activated compiz.
Next to that, simple things as scrolling or moving windows around was really slow (now I'm talking about metacity). The vesa driver gives a smoother experience IMO.

So there are 2 options: either I didn't do something I should've done (in which case these (http://xorg.freedesktop.org/wiki/radeonhd:r6xx_r7xx_branch) directions are incomplete) or they just don't work satisfactorily yet with R7xx hardware (it _is_ marked experimental after all).

[bridgman]

JeanPaul, Compiz uses OpenGL for compositing and that will be running software rendering right now (ie don't bother).

When you say performance was abysmal, which kind of operations are you talking about ? If they're anything but 3D (which isn't accelerated yet) can you post X log and dmesg output ?

[rahman.duran]

I have an X2 4200+ and can play MPEG-2 1080i/1080p video just fine using xv without framedropping (my GPU is a Radeon HD 3850 using fglrx 8.12.) I wonder what the issue is in your system. Where I get into trouble is playing 1080p H.264/MPEG-4 stuff.

How about trying h264 mpeg4-avc profiles :) mpeg2 isn't that much complex for decent x2 CPUs. I get frame drops even if I overclock CPU to 2.6 Ghz.

[rahman.duran]

Did that, installed both the drm driver and the xserver driver, activated radeonhd with the xorg.conf posted earlier...
The performance was abysmal.:D / :(
It would give some empty screen on top of the normal desktop if I ever activated compiz.
Next to that, simple things as scrolling or moving windows around was really slow (now I'm talking about metacity). The vesa driver gives a smoother experience IMO.

So there are 2 options: either I didn't do something I should've done (in which case these (http://xorg.freedesktop.org/wiki/radeonhd:r6xx_r7xx_branch) directions are incomplete) or they just don't work satisfactorily yet with R7xx hardware (it _is_ marked experimental after all).

Are you sure you copy both radeon.ko and drm.ko to proper places? scrolling is fast enough on my hd3200 igp.

[JeanPaul145]

JeanPaul, Compiz uses OpenGL for compositing and that will be running software rendering right now (ie don't bother).

When you say performance was abysmal, which kind of operations are you talking about ? If they're anything but 3D (which isn't accelerated yet) can you post X log and dmesg output ?

I'm talking about simple basic operations like moving a program window or scrolling through something (like a website or a list of contacts, both work far from smooth).

Here are the Xorg.0.log (http://pastebin.com/m6b9deeda) and the dmesg (http://pastebin.com/m27ee3740) output.
I've taken both snapshots of these logs while having fglrx uninstalled and my xorg.conf set to

Section "Device"
Identifier "Configured Video Device"
Driver "radeonhd"
Option "AccelMethod" "exa" # default shadowfb
Option "DRI" "on"
EndSection


@rahman.duran:
Yeah, I just did the whole process again, just to be sure I didn't miss anything on this page (http://xorg.freedesktop.org/wiki/radeonhd:r6xx_r7xx_branch). But I followed those instructions to the letter.
You should note that I have a R7xx gfx card, for which the support is still quite incomplete, and I imagine that the HD3200 igp uses the R6xx driver. That might make a huge difference.

[rahman.duran]

I'm talking about simple basic operations like moving a program window or scrolling through something (like a website or a list of contacts, both work far from smooth).

Here are the Xorg.0.log (http://pastebin.com/m6b9deeda) and the dmesg (http://pastebin.com/m27ee3740) output.
I've taken both snapshots of these logs while having fglrx uninstalled and my xorg.conf set to

Section "Device"
Identifier "Configured Video Device"
Driver "radeonhd"
Option "AccelMethod" "exa" # default shadowfb
Option "DRI" "on"
EndSection


So drm module fails to load. be sure you compile r6xx-r7xx brunch of mesa/drm and be sure you copied drm.ko and radeon.ko to right places. It depends on distros. So you have to overwrite the existing ones.

[susikala]

I'm talking about simple basic operations like moving a program window or scrolling through something (like a website or a list of contacts, both work far from smooth).

Here are the Xorg.0.log (http://pastebin.com/m6b9deeda) and the dmesg (http://pastebin.com/m27ee3740) output.
I've taken both snapshots of these logs while having fglrx uninstalled and my xorg.conf set to

Section "Device"
Identifier "Configured Video Device"
Driver "radeonhd"
Option "AccelMethod" "exa" # default shadowfb
Option "DRI" "on"
EndSection


@rahman.duran:
Yeah, I just did the whole process again, just to be sure I didn't miss anything on [url=http://xorg.freedesktop.org/wiki/radeonhd:r6xx_r7xx_branch]this page[/url. But I followed those instructions to the letter.
You should note that I have a R7xx gfx card, for which the support is still quite incomplete, and I imagine that the HD3200 igp uses the R6xx driver. That might make a huge difference.

Weird. I have the same IGP as you. Have you copied drm.ko and radeon.ko to the right places? On Ubuntu, it should be:

/lib/modules/kernel-version/kernel/drivers/gpu/drm/drm.ko
and
/lib/modules/kernel-version/kernel/drivers/gpu/drm/radeon/radeon.ko

check with lsmod if the drm module is really inserted and whether it's used by radeon (module).

[JeanPaul145]

So drm module fails to load. be sure you compile r6xx-r7xx brunch of mesa/drm and be sure you copied drm.ko and radeon.ko to right places. It depends on distros. So you have to overwrite the existing ones.

Ah. That might be the problem, as I actually mkdirred a drm dir in /lib/modules/2.6.27-11-generic/kernel/drivers/char/.
OK, I'll try and fix that, and let you guys know what happens.

EDIT: it is a big difference indeed!
Scrolling works as smoothly as with fglrx now. nice!
I can't wait to use this fulltime!
Right now the main thing is working compiz with radeonhd, so I guess that would mean waiting for 3D support.
Maybe someone can tell me: is 3D acceleration a must for compiz as a whole, or just for certain plugins? I ask because most (if not all) of the plugins I use look fairly 2D to me (think fadein and fadeout, and although not a compiz plugin, I also want/need AWN support).

Also, I can't really test how radeonhd holds its own against fglrx in fullscreen video until compiz support is in a working state, since that was my problem in the first place - the whole gdm/X/fglrx/Mesa stack locks up if I forget to turn off compiz when watching a video.

[elsie]

I *think* the "Unredirect Fullscreen Windows" option should let the video accel code go straight to the front buffer and skip the extra overhead.

OMG! "Unredirect..." has been enabled for sooo long I completely forgot about it!
I unchecked it & now have stable flicker free XV playback!

My only trouble now is X going blank whenever I switch from battery to AC. I can CTRL-ALT into console & then back to X & problem is resolved...any ideas anyone?

[rafi]

Same problem here, Powerplay seems to be broken in this release. I have a x1600 mobility but aticonfig say my hardware doesn't support powerplay. And just like you or Oyvind I have a blank screen when I switch form battery to AC (the auto-powerstate is still enabled??).

So I'm back to 9-1 and I hope this will be fixed in the Catalyst 9-3.

I'm using Kubuntu 8.10 x86

[elsie]

Same problem here, Powerplay seems to be broken in this release. I have a x1600 mobility but aticonfig say my hardware doesn't support powerplay. And just like you or Oyvind I have a blank screen when I switch form battery to AC (the auto-powerstate is still enabled??).

It seems aticonfig --auto-powerstates=off solves the problem for now.

[rahman.duran]

Ah. That might be the problem, as I actually mkdirred a drm dir in /lib/modules/2.6.27-11-generic/kernel/drivers/char/.
OK, I'll try and fix that, and let you guys know what happens.

EDIT: it is a big difference indeed!
Scrolling works as smoothly as with fglrx now. nice!
I can't wait to use this fulltime!
Right now the main thing is working compiz with radeonhd, so I guess that would mean waiting for 3D support.
Maybe someone can tell me: is 3D acceleration a must for compiz as a whole, or just for certain plugins? I ask because most (if not all) of the plugins I use look fairly 2D to me (think fadein and fadeout, and although not a compiz plugin, I also want/need AWN support).

Also, I can't really test how radeonhd holds its own against fglrx in fullscreen video until compiz support is in a working state, since that was my problem in the first place - the whole gdm/X/fglrx/Mesa stack locks up if I forget to turn off compiz when watching a video.

Well, we have gone a lot off topic :) If you want to use compiz composite then you have to wait for 3D support. It is not about the effects dimension:) but it is about how compositing manager does its job; via OpenGL or Xrender. compiz only supports opengl as I know. But kde 4.x compositing supports both opengl and xrender. So you can use composite effects on kde 4.x with xrender option (as we have nice exa accel for our radeonhd :) ).

[bridgman]

If you're running Gnome you might want to give Metacity a try.

[Boerkel]

Weird. I have the same IGP as you. Have you copied drm.ko and radeon.ko to the right places? On Ubuntu, it should be:

/lib/modules/kernel-version/kernel/drivers/gpu/drm/drm.ko
and
/lib/modules/kernel-version/kernel/drivers/gpu/drm/radeon/radeon.ko

check with lsmod if the drm module is really inserted and whether it's used by radeon (module).

Sorry for more offtopic, but i tried radeonhd as well and i wasn't successful. I'm using Ubuntu Hardy, but i don't have a directory named 'gpu' in /lib/modules/kernel-version/kernel/drivers .

[bridgman]

I think Hardy has the older drm layout; the gpu folders were added fairly recently. Look for an existing file called radeon.ko ; that's the folder you want.

[homerhomer]

This release breaks suspend on my laptop. I went back to 9.1 and everything is okay

bug already listed in the zilla
http://ati.cchtml.com/show_bug.cgi?id=1454

[smagic]

no uvd, powerplay still broken.

I have just sold my ati/amd m780g based hp laptop, and at this point I am happier than when I owned one :).

generally it would hang at the first sight of 3d even in vista, decode h.264 with a lot of stutter, and in linux eat up battery in 1,5 hours with very light usage, hang every now and then, mess up the video, refuse to load at some point and I could go on for hours.
General disappointment.

But that does not concern me anymore because I have sold it! hooray!

[susikala]

no uvd, powerplay still broken.

I have just sold my ati/amd m780g based hp laptop, and at this point I am happier than when I owned one :).

generally it would hang at the first sight of 3d even in vista, decode h.264 with a lot of stutter, and in linux eat up battery in 1,5 hours with very light usage, hang every now and then, mess up the video, refuse to load at some point and I could go on for hours.
General disappointment.

But that does not concern me anymore because I have sold it! hooray!

As said, 780g is very well supported by the FOSS drivers. 3D is soon to come. If you use Linux, you should have some amount of patience with these things in general, I think -- why sell the thing three months after the specs for your GPU were released?

[Boerkel]

As said, 780g is very well supported by the FOSS drivers. 3D is soon to come. If you use Linux, you should have some amount of patience with these things in general, I think -- why sell the thing three months after the specs for your GPU were released?

I hate this "soon" everyone is talking about...

[susikala]

I hate this "soon" everyone is talking about...

AMD/ATI is quite obviously investing more per user in Linux than in any other platform, both with Catalyst and the FOSS drivers. You can only give them a kudos for that and for not folding up linuxwise. The timeframe in which the docs were posted was reasonable, and what we _now_ already have working with the FOSS drivers is a lot more than reasonable. I actually thought they'd need ~half a year to implement proper video playback. Add to it the fact it wouldn't change a wee bit for most personal computer users if AMD didn't develop drivers at all for Linux, and you see you have slight to no reason to complain.

[oyvind]

OMG! "Unredirect..." has been enabled for sooo long I completely forgot about it!
I unchecked it & now have stable flicker free XV playback!

My only trouble now is X going blank whenever I switch from battery to AC. I can CTRL-ALT into console & then back to X & problem is resolved...any ideas anyone?

If you don't care about your GPU clocking down and using less power when on battery, try this:
1) $ aticonfig --auto-powerstates=off
2) Logut/login
3) Pull AC power chord.
4) Put AC power chord back in.

For me, fglrx auto-powerstates actually hard-crashes X, instead of it just going black, like in your case.

--
ATI X1400, Ubuntu 8.10

[benmoran]

Besides the well known issues and not-yet-implemented features, The last two FGLRX releases have been awesome for me.
(Dell Inspiron 1501 w/ Radeon Xpress 1150)
Perhaps my cheap card is less finicky, or maybe i'm lucky. But i've had no problems at all. Each release works better than the last for me.

Just wanted to say thanks to Bridgman and everyone else involved in the Linux side over at AMD. And of course to all the open source devs as well. Im looking forward to switching over to that eventually.

[oyvind]

Would I be right in my suspicion that Textured2D (or similar) is now enabled by default, also for my old and crappy ATI X1400 mobile ? I'm seeing "Textured2D"-like behaviour which I can only remember seeing before when I tested it manually with older Catalyst-releases:
* Certain small tearing patterns, not very annoying though, but easily visible when for instance GDM loads.
* Also slightly laggy/strange repaint-behaviour when moving windows slowly but with constant speed over a window behind it, for instance over a gnome-terminal with old-style transparent background.

2D perf is overall good, though. And I'm not talking about Compiz-env here, just plain Metacity.

[hpestilence]

Why does fglrx x86_64 driver hardcode /usr/lib64 in it's binary? It messes up with distros providing a pure 64-bit Linux. For example, Arch Linux http://bbs.archlinux.org/viewtopic.php?id=65960

It's been a problem since catalyst 9.1

[RealNC]

As if a link lib->lib64 is going to bring the world down. Seriously.

[piccolotux]

With my ati mobility 1600 powerplay stopped working. So I sadly have to confirm the powerplay bug. Back to 9.1.

[NeoBrain]

As if a link lib->lib64 is going to bring the world down. Seriously.

Well, if it doesn't work out of the box, people may not be able to install the driver at all.
There are /some/ newbies using Linux after all...

[BlackStar]

As if a link lib->lib64 is going to bring the world down. Seriously.
Obviously not, but this is a regression and should be fixed.

Besides, it's annoying as it clashes with the policy of the distribution (pure 64-bit, with 64-bit libraries in /usr/lib).

[bugmenot]

As if a link lib->lib64 is going to bring the world down. Seriously.

the common linux dir hierarchy (eg, see http://en.wikipedia.org/wiki/Filesystem_Hierarchy_Standard) is not intended to support multilib.
further, xorg and its components are all installed under /lib or /usr/lib.
so since this should be a _linux_ not a debian/ubuntu/gentoo/some-other-multilib-distro driver, this is a major regression!
btw, it's the dumbest idea to hardcode paths in apps!
hope they change this in upcoming versions.

[PuckPoltergeist]

the common linux dir hierarchy (eg, see http://en.wikipedia.org/wiki/Filesystem_Hierarchy_Standard) is not intended to support multilib.

Did you read the standard? /lib is for the native binary format. Most (all?) rpm distros don't follow this and put the 32bit lib into /lib. They break the standard there. Gentoo does it right, putting 32bit libs to /lib32 and 64bit libs into /lib64. /lib is only a symlink to the native format.
So, multilib is supported by linux but not right implemented by most distros.

[MU_Engineer]

Did you read the standard? /lib is for the native binary format. Most (all?) rpm distros don't follow this and put the 32bit lib into /lib. They break the standard there. Gentoo does it right, putting 32bit libs to /lib32 and 64bit libs into /lib64. /lib is only a symlink to the native format.
So, multilib is supported by linux but not right implemented by most distros.

Debian does it the same way as Gentoo as there is /lib32 and /lib64. /lib links to /lib64 /lib32 is linked to /emul/ia32-linux/lib.

[RealNC]

the common linux dir hierarchy (eg, see http://en.wikipedia.org/wiki/Filesystem_Hierarchy_Standard) is not intended to support multilib.
further, xorg and its components are all installed under /lib or /usr/lib.
so since this should be a _linux_ not a debian/ubuntu/gentoo/some-other-multilib-distro driver, this is a major regression!


The problem is, that it *is* a debian/ubuntu/etc driver. Only certain distros are supported by the driver. It's stated in the release notes.

Of course hardcoding paths is brain damage, but still, the driver package is only for specific distros. The rest must somehow manage to comply with its expectations.

[Kano]

Well best it to use packageing of course. If you use the pure installer then I may fail. The Debian packageing is just too often with errors, so never rely on it.

[BlackStar]

Can anyone with an ati 2x00+ or nvidia 8x00+ run the following program? It tries to create an OpenGL 3.0 context and render some colors to a small window - and, despite my best efforts, it fails unceremoniously on my 4850 with 9.2 on x86_64.


wget http://www.opentk.com/files/issues/gl3.cpp
g++ gl3.cpp -o gl3 -lX11 -lGL
./gl3


Does it work for you? What hardware? What drivers? x86 or x86_64?

[MU_Engineer]

Can anyone with an ati 2x00+ or nvidia 8x00+ run the following program? It tries to create an OpenGL 3.0 context and render some colors to a small window - and, despite my best efforts, it fails unceremoniously on my 4850 with 9.2 on x86_64.


wget http://www.opentk.com/files/issues/gl3.cpp
g++ gl3.cpp -o gl3 -lX11 -lGL
./gl3


Does it work for you? What hardware? What drivers? x86 or x86_64?

It does not work for me:

./gl3
GLX_ARB_get_proc_address GLX_ARB_multisample GLX_EXT_import_context GLX_EXT_visual_info GLX_EXT_visual_rating GLX_MESA_swap_control GLX_OML_swap_method GLX_OML_sync_control GLX_SGI_video_sync GLX_SGIS_multisample GLX_SGIX_fbconfig GLX_SGIX_visual_select_group
glXCreateContextAttribsARB entry point not found


GPU: Radeon HD 3850
Drivers: Catalyst 8.12
Distro: Debian "Squeeze" amd64

[Kano]

@BlackStar

A bit short but it runs fine with nvidia 8800 gts 512.

GLX_EXT_visual_info GLX_EXT_visual_rating GLX_SGIX_fbconfig GLX_SGIX_pbuffer GLX_SGI_video_sync GLX_SGI_swap_control GLX_EXT_texture_from_pixmap GLX_ARB_create_context GLX_ARB_multisample GLX_NV_float_buffer GLX_ARB_fbconfig_float GLX_EXT_framebuffer_sRGB GLX_ARB_get_proc_address
Getting framebuffer config
Getting XVisualInfo
Creating colormap
Creating window
Mapping window
Creating context
Making context current

[BlackStar]

Thanks for testing!

Ati officially introduced GL3 support with catalyst 9.1. There was a chance it would work on 8.12, but it seems the relevant entry points are not available on this version (GLX_ARB_create_context is missing from the extension list).

Edit: missed your post Kano. So it seems to work fine on nvidia, which probably indicates a driver bug on ati's side.

Anyone with an ati card and catalyst 9.1 or 9.2 care to make a try?

[hpestilence]

GLX_ARB_create_context GLX_ARB_get_proc_address GLX_ARB_multisample GLX_EXT_import_context GLX_EXT_visual_info GLX_EXT_visual_rating GLX_MESA_swap_control GLX_NV_swap_group GLX_OML_swap_method GLX_SGI_video_sync GLX_SGIS_multisample GLX_SGIX_fbconfig GLX_SGIX_swap_barrier GLX_SGIX_swap_group GLX_SGIX_visual_select_group
Getting framebuffer config
Getting XVisualInfo
Creating colormap
Creating window
Mapping window
Creating context
Segmentation fault

My card is an atihd4850 with the 9.2 catalyst driver. Arch Linux x86_64.

[BlackStar]

Beautiful, that's what I'm seeing too. Not too surprising, considering I'm also running Arch Linux x86_64 with a 4850 and the 9.2 driver :D

[MU_Engineer]

Thanks for testing!

Ati officially introduced GL3 support with catalyst 9.1. There was a chance it would work on 8.12, but it seems the relevant entry points are not available on this version (GLX_ARB_create_context is missing from the extension list).

Edit: missed your post Kano. So it seems to work fine on nvidia, which probably indicates a driver bug on ati's side.

Anyone with an ati card and catalyst 9.1 or 9.2 care to make a try?

I suppose I'll have to wait until 9.1 or 9.2 hits Debian testing as installing the packages from the fglrx installer often causes a lot of problems.

[PuckPoltergeist]

GLX_ARB_create_context GLX_ARB_get_proc_address GLX_ARB_multisample GLX_EXT_import_context GLX_EXT_visual_info GLX_EXT_visual_rating GLX_MESA_swap_control GLX_NV_swap_group GLX_OML_swap_method GLX_SGI_video_sync GLX_SGIS_multisample GLX_SGIX_fbconfig GLX_SGIX_swap_barrier GLX_SGIX_swap_group GLX_SGIX_visual_select_group
Getting framebuffer config
Getting XVisualInfo
Creating colormap
Creating window
Mapping window
Creating context
Segmentation fault

My card is an atihd4850 with the 9.2 catalyst driver. Arch Linux x86_64.

here too, Gentoo x86_64, HD3650 AGP, Catalyst-9.2

[lbnbfhytoydkdd]

For me, 9.2 installs and runs under 64bit openSUSE 11.0 with a 4850 card and compositing under KDE 4 is working, too. However, any program that tries to read OpenGL information just hangs, with X using ~30% of processor time on an AMD Quadcore processor. The programs I tried are fglrxinfo, amdcccle, fgl_glxgears, KDE4 compositing settings, and games under Wine. Once any of these apps are started, they hang forever. The system itself keeps working fine however.
Reverting back to 9.1 resolves this. Looks like a regression to me.

[mirv]

For me, 9.2 installs and runs under 64bit openSUSE 11.0 with a 4850 card and compositing under KDE 4 is working, too. However, any program that tries to read OpenGL information just hangs, with X using ~30% of processor time on an AMD Quadcore processor. The programs I tried are fglrxinfo, amdcccle, fgl_glxgears, KDE4 compositing settings, and games under Wine. Once any of these apps are started, they hang forever. The system itself keeps working fine however.
Reverting back to 9.1 resolves this. Looks like a regression to me.

is there any info from dmesg? I've had similar things before, typically related to an incomplete / incorrect driver installation.

[Chewi]

I've been waiting for this release all month but I was out on Friday so I didn't notice this till now! I hoped it would build against 2.6.29. I haven't tried yet but I don't think it will. It has indeed fixed the problem in MythTV though, that had been bugging me for ages. GUI performance in MythTV seems much better in general.

[Boerkel]

Has anyone an idea why i can't use the xinerama-option in the xorg.conf anymore. I'm using Ubuntu Hardy with a Mobility Radeon HD 2600 and it worked quite well with 8.12 and 9.1, but when i'm trying to set this option in 9.2 the screen is just black and nothing works anymore.
Is there anyone who could help me?

[lbnbfhytoydkdd]

is there any info from dmesg? I've had similar things before, typically related to an incomplete / incorrect driver installation.

This is the fglrx output in dmesg:
fglrx: module license 'Proprietary. (C) 2002 - ATI Technologies, Starnberg, GERMANY' taints kernel.
[fglrx] Maximum main memory to use for locked dma buffers: 1885 MBytes.
[fglrx] vendor: 1002 device: 9442 count: 1
[fglrx] ioport: bar 4, base 0xc000, size: 0x100
[fglrx] Driver built-in PAT support is enabled successfully
[fglrx] module loaded - fglrx 8.58.2 [Feb 4 2009] with 1 minors
[fglrx] Firegl kernel thread PID: 2555
[fglrx] Gart USWC size:942 M.
[fglrx] Gart cacheable size:60 M.
[fglrx] Reserved FB block: Shared offset:0, size:1000000
[fglrx] Reserved FB block: Unshared offset:fdff000, size:201000
[fglrx] Reserved FB block: Unshared offset:1fffb000, size:5000

Nothing interesting I would say. I also had aticonfig configure my xorg.conf again, just to be on the safe side, but anything trying to read the OpenGL information still hangs.

Does anyone have other ideas? :confused:

[Chewi]

I know that if the user can't read xorg.conf then things break so check that.

[lbnbfhytoydkdd]

xorg.conf is perfectly readable when logged in as a normal user.

[lup0]

When upgrading from catalyst 9.1 to 9.2 HDMI audio stopped working for me. Reverting back to 9.1 makes the audio work again. Is there some way to debug this problem? aplay -l lists the HDA ATI HDMI in the same card and device location with both drivers.

[lordmozilla]

I still have the same problem as on 9.1, When I put compiz on, xfce (ubuntu 8.10), And i play a video, and make it full screen with vlc or smplayer desktop crashes irreversibly. ssh goes down etc...

However compiz performance is top notch absolutely no problems

If I disable compiz and emerald everything is fine.

However video tearing is completely erradicated so the bottom 20% of my video window is not out of sync all the time, which is nice.

It's a bit of a shame but oh well, life is alot better than previous in the fglrx world.

Using ATI HD4850 1gb GDDR3

[RealNC]

9.2 results in non-working KDE4 + composite. "Non working" meaning that it *seems* to work, but is extremely slow. 1 minute to login to KDE, 1 minute to open a menu, etc. Everything is practically frozen except for the mouse cursor.

I wonder when all this breakage will end. I suppose it's intentional. Breaking things makes the next release have entries in the changelog which otherwise would stay empty :D

[bridgman]

9.2 results in non-working KDE4 + composite. "Non working" meaning that it *seems* to work, but is extremely slow. 1 minute to login to KDE, 1 minute to open a menu, etc. Everything is practically frozen except for the mouse cursor.

I wonder when all this breakage will end. I suppose it's intentional. Breaking things makes the next release have entries in the changelog which otherwise would stay empty :D

Are you saying that 9.2 is slower than 9.1 when running KDE4 with composite ? I hadn't heard anything about a regression, just that it was slow.

[jonnycat26]

9.2 results in non-working KDE4 + composite. "Non working" meaning that it *seems* to work, but is extremely slow. 1 minute to login to KDE, 1 minute to open a menu, etc. Everything is practically frozen except for the mouse cursor.


For me, 9.2 broke KDE4 + composite entirely, that is, it broke it entirely on those rare occasions when I could convince X to startup properly.

I would get hard crashes when I would restore applications from the system tray or launch applications. I think it may have been the fade in/fade out effect that it didn't like, but I didn't really get to test too much as it frustrated the hell out of me.

For now, my ATI card is in the closet and I'm back to NVidia. If a future catalyst release fixes all of these issues, I may switch back. No offense to bridgeman, I think it's great that you're here to listen and help out. But your comments make it seem like things aren't going to get better in the immediate future, so for now, I've got to leave the Radeon behind.

[RealNC]

Are you saying that 9.2 is slower than 9.1 when running KDE4 with composite ? I hadn't heard anything about a regression, just that it was slow.

Well, clicking on something and having to wait for about a minute for the result is not exactly "slow". :D

[bridgman]

OK, call it glacial then :D

Are you saying that the 9.2 release is *more* glacial than 9.1 ?

[NSLW]

I had the opportunity to test Catalyst 9.2 on Fedora 10 i386 (GNOME) -installed from rpmfusion testing updates repo- with ATI Mobility X1600 and here is what worked (and not) after about 10 min of using it:

- tear free movies playback (Xv-AVIVO) with metacity but not with compiz fusion (v.0.7.8-2),
- blank screen after watching in fullscreen and then windowed mode SMPlayer (0.6.6 (SVN r2599)) when compiz fusion enabled and of course I had to make hard reset,
- lm_sensors (v.3.0.2-1) doesn't detect on-board temp sensor
- aticonfig says: card doesn't support powerplay (How it's possible with laptop gfx?)
- aticonfig displayed some error when I wanted to check on-board sensor temperature (card has got such a sensor in CCC under WindowsXP)

[RealNC]

OK, call it glacial then :D

Are you saying that the 9.2 release is *more* glacial than 9.1 ?

9.1 works just as "well" as any other previous release. 9.2 is bugged here. It's not about "slow" it's about some bug that results in sub-1FPS performance in everything.

Also, totally out of the blue, some elements in KDE4 appear differently than with 9.1. For example, the desktop buttons on the task bar are laid out horizontally instead of vertically. I was not able to check for other weird stuff because you can't do anything on the desktop; it's like looking at a screenshot of a desktop or trying to use true-color, uncompressed VNC over a 8kbbs phone modem connection.

There was nothing helpful in Xorg.0.log nor dmesg. Xorg.0.log also gets spammed to death with modeline messages.

I'm on Gentoo AMD64*, kernel 2.6.28.5 and X.Org 1.5.3. Radeon HD4870 512MB.

*And because of this you never know if it's a packaging bug or not. I don't think the installer works on Gentoo; we rely on the Gentoo fglrx maintainer to package it.

[bridgman]

OK, thanks.

[jonnycat26]

9.1 works just as "well" as any other previous release. 9.2 is bugged here. It's not about "slow" it's about some bug that results in sub-1FPS performance in everything.

Also, totally out of the blue, some elements in KDE4 appear differently than with 9.1. For example, the desktop buttons on the task bar are laid out horizontally instead of vertically. I was not able to check for other weird stuff because you can't do anything on the desktop; it's like looking at a screenshot of a desktop or trying to use true-color, uncompressed VNC over a 8kbbs phone modem connection.

There was nothing helpful in Xorg.0.log nor dmesg. Xorg.0.log also gets spammed to death with modeline messages.

Try backing up your .kde4 directory and starting fresh. If that works, restore your .kde4 directory and start selectively deleting settings (kwin may help).

[RealNC]

I'd rather keep it as-is because there's absolutely nothing in 9.2 that makes it worth the upgrade.