The AMD/ATI Bar & Grill

h00pertr0n · 2013-11-27 19:03:25

Hi, major Arch noob here.

I have run into problems installing catalyst on my Arch/Gnome installation (currently using the open source driver). I follow the installation/configuration procedures on the wiki, but gnome does not boot, instead I get "oh no something went wrong etc....".

I gather this could be because some of my packages are too up to date for catalyst to handle. Is this so and if so how do I check?

Thanks

clfarron4 · 2013-11-27 19:22:26

h00pertr0n wrote:

Hi, major Arch noob here.
I have run into problems installing catalyst on my Arch/Gnome installation (currently using the open source driver). I follow the installation/configuration procedures on the wiki, but gnome does not boot, instead I get "oh no something went wrong etc....".
I gather this could be because some of my packages are too up to date for catalyst to handle. Is this so and if so how do I check?
Thanks

GNOME Shell doesn't work for me either with AMD Catalyst (it does on my Intel Graphics setup though). That said, I have no interest in getting GNOME Shell to work because I use Cinnamon (and a variant I like to call awesome-cinnamon).

Last edited by clfarron4 (2013-11-27 19:23:54)

Kuci · 2013-11-27 19:30:40

GNOME works with the latest AMD Catalyst for me. But I needed to generate new xorg.conf using:

$ aticonfig --adapter=all --initial

h00pertr0n · 2013-11-27 19:38:23

Just tried again using the wiki instructions. All working fine now - either something got updated since yesterday or I didn't follow one of the steps properly before.

lano1106 · 2013-11-27 19:52:21

ivdok wrote:

lano1106 wrote:
I can check how mplayer behaves with Xvba (if they still have Xvba support in recent mplayer releases...) but do not hold your breathe
Yes-yes, if it's not hard for you

Vi0L0 has been faster than me. mplayer turns out to be decoding fine by just ignoring the X error. I'll create a patch for libva-xvba-driver to emulate this behavior for every application using this driver.

Concerning the mplayer native Xvba support, the wiki page is not quite accurate. By reading the phoronix article pointed by the wiki page, you end up understanding that libXVBAW.so (What does the 'W' stands for? XvBA Wrapper?) expose the XvMC API and you can recompile mplayer to let it use the catalyst implementation of XvMC. If you ask me, this is not really pertinent today since XvMC helps MPEG2 decoding.

To be useful hw decoding should support H.264 and MPEG4.

dfanz0r · 2013-11-27 20:15:33

So i cannot seem to get my R9 290 to work.
xorg log: http://pastebin.com/uWKbqDYb

Has anyone actually gotten it to work. I am using Vi0L0's catalyst-test from the aur.
Everything works if i use one of my other radeon graphics cards.

lano1106 · 2013-11-27 21:01:46

Vi0L0 wrote:

xvba/vaapi is working fine with newest catalyst (13.11 beta 9.4) and xorg-server 1.13.4. I have used [xorg113] repo. Both cvlc and mplayer are working without an error here. Can someone confirm?
[xorg113]
Server = http://catalyst.wirephire.com/repo/xorg113/$arch
I'm typically downgrading 3 packages:
pacman -S xorg-server xorg-server-common xf86-input-evdev

Vi0L0, I have tested with Xserver 1.13.4. I have no more BadRequest error. Now I have a BadValue error :-)

I have added the traces at:

http://ati.cchtml.com/show_bug.cgi?id=950

Maybe the HW revision influence the result. I am doing my tests with a HD 7970.

ivdok · 2013-11-28 10:20:33

lano1106 wrote:

ivdok wrote:
lano1106 wrote:
I can check how mplayer behaves with Xvba (if they still have Xvba support in recent mplayer releases...) but do not hold your breathe
Yes-yes, if it's not hard for you
Vi0L0 has been faster than me. mplayer turns out to be decoding fine by just ignoring the X error. I'll create a patch for libva-xvba-driver to emulate this behavior for every application using this driver.
Concerning the mplayer native Xvba support, the wiki page is not quite accurate. By reading the phoronix article pointed by the wiki page, you end up understanding that libXVBAW.so (What does the 'W' stands for? XvBA Wrapper?) expose the XvMC API and you can recompile mplayer to let it use the catalyst implementation of XvMC. If you ask me, this is not really pertinent today since XvMC helps MPEG2 decoding.
To be useful hw decoding should support H.264 and MPEG4.

lano1106 wrote:

Vi0L0 wrote:
xvba/vaapi is working fine with newest catalyst (13.11 beta 9.4) and xorg-server 1.13.4. I have used [xorg113] repo. Both cvlc and mplayer are working without an error here. Can someone confirm?
[xorg113]
Server = http://catalyst.wirephire.com/repo/xorg113/$arch
I'm typically downgrading 3 packages:
pacman -S xorg-server xorg-server-common xf86-input-evdev
Vi0L0, I have tested with Xserver 1.13.4. I have no more BadRequest error. Now I have a BadValue error :-)
I have added the traces at:
http://ati.cchtml.com/show_bug.cgi?id=950
Maybe the HW revision influence the result. I am doing my tests with a HD 7970.

Guys! I've found the culprit - oh, and the latest drivers + XvBA-VAAPI are working again
In the morning, when I was in the bus, i was reading my e-mail inbox, and have been reading AMD newsletter regarding your XvBA. In all your tickets I found that you have been using Xorg 1.13 all along. "But why not 1.14?", I thought. "Catalyst ≥ 13.6 supports 1.14! Wait. They have working XvBA-VAAPI. I have only with old Xorg and drivers. They use newer drivers. But not Xorg!" Later, approx. half a minute ago, I have rechecked my hyphotesis about old Xorg - and I have won! Seems what FreeDesktop guys like to break standarts that they implement themselves

Some useless logs from terminal:

[ivdok@localhost ~]$ Xorg -version
X.Org X Server 1.13.4
Release Date: 2013-04-17
X Protocol Version 11, Revision 0
Build Operating System: Linux 3.8.10-1-ARCH i686
Current Operating System: Linux localhost 3.12.1-3-ARCH #1 SMP PREEMPT Tue Nov 26 11:40:55 CET 2013 i686
Kernel command line: BOOT_IMAGE=/boot/vmlinuz-linux root=UUID=691f4e0c-c7fc-40c5-bf00-e8ed2f0e4392 rw nomodeset quiet
Build Date: 28 April 2013 01:23:21PM
Current version of pixman: 0.32.4
Before reporting problems, check http://wiki.x.org
to make sure that you have the latest version.
[ivdok@localhost ~]$ vainfo
libva info: VA-API version 0.34.0
libva info: va_getDriverName() returns 0
libva info: User requested driver 'xvba'
libva info: Trying to open /usr/lib/dri//xvba_drv_video.so
libva info: Found init function __vaDriverInit_0_32
libva info: va_openDriver() returns 0
vainfo: VA-API version: 0.34 (libva 1.2.1)
vainfo: Driver version: Splitted-Desktop Systems XvBA backend for VA-API - 0.7.8
vainfo: Supported profile and entrypoints
VAProfileH264High : VAEntrypointVLD
VAProfileVC1Advanced : VAEntrypointVLD
[ivdok@localhost ~]$ vlc
VLC media player 2.1.1 Rincewind (revision 2.1.0-207-g89c9520)
[0x8490568] pulse audio output error: PulseAudio server connection failure: Connection refused
[0x83fe910] main libvlc: Running vlc with the default interface. Use 'cvlc' to use vlc without interface.
Failed to open VDPAU backend libvdpau_nvidia.so: cannot open shared object file: No such file or directory
[0xb120d430] vdpau generic error: device creation failure: error 1
libva info: VA-API version 0.34.0
libva info: va_getDriverName() returns 0
libva info: User requested driver 'xvba'
libva info: Trying to open /usr/lib/dri//xvba_drv_video.so
libva info: Found init function __vaDriverInit_0_32
No accelerated IMDCT transform found
libva info: va_openDriver() returns 0
xvba_video: XVBA_GetSurface(): status 2
[0x84d4e38] avcodec decoder: Using VA API version 0.34 for hardware decoding.
Failed to open VDPAU backend libvdpau_nvidia.so: cannot open shared object file: No such file or directory
[0xb120d430] vdpau generic error: device creation failure: error 1
libva info: VA-API version 0.34.0
libva info: va_getDriverName() returns 0
libva info: User requested driver 'xvba'
libva info: Trying to open /usr/lib/dri//xvba_drv_video.so
libva info: Found init function __vaDriverInit_0_32
libva info: va_openDriver() returns 0
xvba_video: XVBA_GetSurface(): status 2
[0x84d4e38] avcodec decoder: Using VA API version 0.34 for hardware decoding.
Fontconfig warning: FcPattern object size does not accept value "0"
Fontconfig warning: FcPattern object size does not accept value "0"
Fontconfig warning: FcPattern object size does not accept value "0"
Fontconfig warning: FcPattern object size does not accept value "0"
[0xa094be68] main vout display error: Failed to resize display

Last edited by ivdok (2013-11-28 10:33:59)

Vi0L0 · 2013-11-28 11:01:08

ivdok wrote:

In all your tickets I found that you have been using Xorg 1.13 all along.

Not at all. Both lano1106 and I was using xserver 1.14.4 from the very begining of this case.
Then I switched to xserver 1.13.4 and found out that it's working perfectly for me. As i wrote in bug report 950:

With catalyst 13.11 beta 9.4 + xorg-server 1.13.4 + xvba-driver 0.8.0
both cvlc and mplayer (with vaapi support) are working fine, giving no "X Error of failed request".
Switching to xorg-server 1.14.4 ends up in cvlc being completelly unable to start because of mentioned X Error, but mplayer is able to work with it. It's mentioning that there was a X error, but ignores it and decodes movie normally.

As you can see with xserver 1.14.4 and radeon hd 7850 mplayer is working fine for me (decodes well while ignoring the X Error), but lano1106 wasn't able to get such a positive result with hes hd 7970.
(just noted that i should be more specific and add that cvlc was able to start but not able to decode and draw)

ivdok wrote:

Later, approx. half a minute ago, I have rechecked my hyphotesis about old Xorg - and I have won! Seems what FreeDesktop guys like to break standarts that they implement themselves
Some useless logs from terminal:
[ivdok@localhost ~]$ Xorg -version
X.Org X Server 1.13.4
Release Date: 2013-04-17
X Protocol Version 11, Revision 0
...

So what's the conclusion? Logs are not showing decoding so it's hard to tell anything basing on them.

Last edited by Vi0L0 (2013-11-28 11:03:45)

ivdok · 2013-11-28 11:21:16

Vi0L0 wrote:

Not at all. Both lano1106 and I was using xserver 1.14.4 from the very begining of this case.
Then I switched to xserver 1.13.4 and found out that it's working perfectly for me.

Well, maybe I didn't look entirely everything at once.

Vi0L0 wrote:

As i wrote in bug report 950:
With catalyst 13.11 beta 9.4 + xorg-server 1.13.4 + xvba-driver 0.8.0
both cvlc and mplayer (with vaapi support) are working fine, giving no "X Error of failed request".
Switching to xorg-server 1.14.4 ends up in cvlc being completelly unable to start because of mentioned X Error, but mplayer is able to work with it. It's mentioning that there was a X error, but ignores it and decodes movie normally.
As you can see with xserver 1.14.4 and radeon hd 7850 mplayer is working fine for me (decodes well while ignoring the X Error), but lano1106 wasn't able to get such a positive result with hes hd 7970.

I understand now. 1.14 is not implemented properly and entirely - it's "beta" drivers, after all.

Vi0L0 wrote:

So what's the conclusion? Logs are not showing decoding so it's hard to tell anything basing on them.

It works all the same, as stable, in my opinion - only VLC segfaults on "Breaking Bad" for unknown reason, but I had this issue with stable driver too, and, besides, it's not a nuisance - it just crashes when trying to play next video, that's all. Manual startup from Thunar works as expected.

Vi0L0 · 2013-11-28 14:31:44

dfanz0r wrote:

So i cannot seem to get my R9 290 to work.
xorg log: http://pastebin.com/uWKbqDYb
Has anyone actually gotten it to work. I am using Vi0L0's catalyst-test from the aur.
Everything works if i use one of my other radeon graphics cards.

"<default monitor>" looks strange, especially when used two times with different screens. How many displays you got there? Can you paste xorg.conf?

dfanz0r · 2013-11-29 06:24:40

Vi0L0 wrote:

dfanz0r wrote:
So i cannot seem to get my R9 290 to work.
xorg log: http://pastebin.com/uWKbqDYb
Has anyone actually gotten it to work. I am using Vi0L0's catalyst-test from the aur.
Everything works if i use one of my other radeon graphics cards.
"<default monitor>" looks strange, especially when used two times with different screens. How many displays you got there? Can you paste xorg.conf?

I had 4 plugged in at the time. Currently i am away (took the computer with me lol) and only have 1 plugged in.

Here a second log with only the one monitor connected: http://pastebin.com/mzNdXvJi
Here is the conf used when the above log was generated: http://pastebin.com/19M2z3S9

gee · 2013-11-30 10:28:51

Hey there,

I just built a 2nd machine and started using the fglrx.
I wanted to give a big thanks to Vi0L0 for this repos, they were so very helpful!
I am currently running with a 4670 but soon replacing it with a 79xx model, so for now legacy catalyst and app-stream 2.7.
Using these 2 simple open cl tests fail with kernel compilation failures, is there any known problem with the old catalyst and the old open cl sdk?
I should be getting the new card soon so don't want to waste too much time if it's related to the older software...
On the other hand if it's unrelated I'd rather fix it before

Kuci · 2013-11-30 11:42:24

gee wrote:

Hey there,
I just built a 2nd machine and started using the fglrx.
I wanted to give a big thanks to Vi0L0 for this repos, they were so very helpful!
I am currently running with a 4670 but soon replacing it with a 79xx model, so for now legacy catalyst and app-stream 2.7.
Using these 2 simple open cl tests fail with kernel compilation failures, is there any known problem with the old catalyst and the old open cl sdk?
I should be getting the new card soon so don't want to waste too much time if it's related to the older software...
On the other hand if it's unrelated I'd rather fix it before

4XXX's architecture is not optimized for OpenCL, that's the reason why those cards are not supported anymore.

gee · 2013-12-01 04:41:42

I actually found a command line that worked with cgminer... not very high hash rate though, but I just wanted to make it work.

(-w 64 was the trick)
Yet those cl tests still fail... strange, but maybe they use "advanced" functionalities that the 4XXX do not support or something....
Thanks!

gee · 2013-12-02 04:15:07

oh and yes, with a 7xxx card, no more error on the opencl tests, I guess that was it then...
Thanks!

Vi0L0 · 2013-12-04 20:48:59

FYI Lately on Steam I bought Metro Last Light and it's working well here on hd 7850 and with catalyst 13.11 beta 9.4. When Metro LL was released there was an info that AMD is working on to support it properly, but I never saw a news telling it's fine now. Well... it's fine here ;P
I didn't play for a long time, but everything was rendered correctly and fps was also playable - around 40 to 60 fps on highest settings @ fullhd, so like on windows. I can see that tesselation is missing but AFAIK it's normal: http://www.overclock.net/t/1439965/metr … comparison

dfanz0r wrote:

Vi0L0 wrote:
dfanz0r wrote:
So i cannot seem to get my R9 290 to work.
xorg log: http://pastebin.com/uWKbqDYb
Has anyone actually gotten it to work. I am using Vi0L0's catalyst-test from the aur.
Everything works if i use one of my other radeon graphics cards.
"<default monitor>" looks strange, especially when used two times with different screens. How many displays you got there? Can you paste xorg.conf?
I had 4 plugged in at the time. Currently i am away (took the computer with me lol) and only have 1 plugged in.
Here a second log with only the one monitor connected: http://pastebin.com/mzNdXvJi
Here is the conf used when the above log was generated: http://pastebin.com/19M2z3S9

What if you will connect one display and then generated (fresh) xorg.conf with aticonfig?

joaocandre · 2013-12-04 21:39:22

So I installed catalyst drivers today, on an up to date Arch system, and suddenly Steam is showing an error when logging in:

Steam wrote:

OpenGL GLX context is not using direct rendering, which may cause performance problems.
For more information visit https://support.steampowered.com/kb_art … -EYZB-7457.

Anyone else having this problem? It was OK when I only had the open source drivers. I installed catalyst from VI0L0 repos, had to to install beta since I'm running Xorg 1.14.

According to [ glxinfo | grep "direct rendering" ] I am using direct rendering, so I don't know what Steam is complaining about.

ivdok · 2013-12-05 04:18:29

Vi0L0 wrote:

FYI Lately on Steam I bought Metro Last Light and it's working well here on hd 7850 and with catalyst 13.11 beta 9.4. When Metro LL was released there was an info that AMD is working on to support it properly, but I never saw a news telling it's fine now. Well... it's fine here ;P
I didn't play for a long time, but everything was rendered correctly and fps was also playable - around 40 to 60 fps on highest settings @ fullhd, so like on windows. I can see that tesselation is missing but AFAIK it's normal.

IDK. Running Assassin's Creed II on Wine was rather painful - GPU worked at likely 20% performance, severe lag and audio/script de-sync included. Nvidia reports show what Wine is working well for this particular game. So Linux is not yet gaming platform for everyday, IMO. SteamOS is no exception.

ChemBro · 2013-12-05 05:27:43

ivdok wrote:

Vi0L0 wrote:
FYI Lately on Steam I bought Metro Last Light and it's working well here on hd 7850 and with catalyst 13.11 beta 9.4. When Metro LL was released there was an info that AMD is working on to support it properly, but I never saw a news telling it's fine now. Well... it's fine here ;P
I didn't play for a long time, but everything was rendered correctly and fps was also playable - around 40 to 60 fps on highest settings @ fullhd, so like on windows. I can see that tesselation is missing but AFAIK it's normal.
IDK. Running Assassin's Creed II on Wine was rather painful - GPU worked at likely 20% performance, severe lag and audio/script de-sync included. Nvidia reports show what Wine is working well for this particular game. So Linux is not yet gaming platform for everyday, IMO. SteamOS is no exception.

But you're talking about a game through Wine and we all know Wine uses these special Nvidia GL something to render the game (obviously AMD does not have those). Metro Last Light has a native Linux version, that is something different, especially for AMD cards.

Vi0L0 · 2013-12-05 07:48:59

joaocandre wrote:

So I installed catalyst drivers today, on an up to date Arch system, and suddenly Steam is showing an error when logging in:
Steam wrote:
OpenGL GLX context is not using direct rendering, which may cause performance problems.
For more information visit https://support.steampowered.com/kb_art … -EYZB-7457.
Anyone else having this problem? It was OK when I only had the open source drivers. I installed catalyst from VI0L0 repos, had to to install beta since I'm running Xorg 1.14.
According to [ glxinfo | grep "direct rendering" ] I am using direct rendering, so I don't know what Steam is complaining about.

If you are using x86_64 system then have you installed lib32-catalyst-utils?

ChemBro wrote:

ivdok wrote:
Vi0L0 wrote:
FYI Lately on Steam I bought Metro Last Light and it's working well here on hd 7850 and with catalyst 13.11 beta 9.4. When Metro LL was released there was an info that AMD is working on to support it properly, but I never saw a news telling it's fine now. Well... it's fine here ;P
I didn't play for a long time, but everything was rendered correctly and fps was also playable - around 40 to 60 fps on highest settings @ fullhd, so like on windows. I can see that tesselation is missing but AFAIK it's normal.
IDK. Running Assassin's Creed II on Wine was rather painful - GPU worked at likely 20% performance, severe lag and audio/script de-sync included. Nvidia reports show what Wine is working well for this particular game. So Linux is not yet gaming platform for everyday, IMO. SteamOS is no exception.
But you're talking about a game through Wine and we all know Wine uses these special Nvidia GL something to render the game (obviously AMD does not have those). Metro Last Light has a native Linux version, that is something different, especially for AMD cards.

Also please remember that wine devs are working on nvidias cards, it's probably the reason why wine doesn't support radeons well.
Steam is using prepared by nvidia togl library which is "translating" directx to opengl, but it's designed to work on every (opengl) software and hardware capable system.

joaocandre · 2013-12-05 11:54:42

Vi0L0 wrote:

If you are using x86_64 system then have you installed lib32-catalyst-utils?

Yup, that solved. I somehow missed that line on the wiki. Thanks!

haagch · 2013-12-05 14:18:53

ChemBro wrote:

But you're talking about a game through Wine and we all know Wine uses these special Nvidia GL something to render the game (obviously AMD does not have those).

I don't buy that. Try the open source driver. It's often much better than catalyst in wine.

snack · 2013-12-05 17:25:30

I'm beginning to toy with OpenCL. My development laptop has a Mobility Radeon HD3470 (with catalyst-legacy) which is not OpenCL capable, but for my early experiments I'm fine with using the CPU (Intel P8400) as OpenCL device. I've installed APP-SDK 2.9 and opencl-headers, and now I'm trying to compile a simple example taken from the book "OpenCL programming guide". The compilation has no problem, but when I launch it it gives a segfault on the first OpenCL instruction it tries to execute:

errNum = clGetPlatformIDs(0, NULL, &numPlatforms);

I don't know what might cause this problem. Maybe APP-SDK 2.9 is not compatible with Catalyst 13.1 (the current legacy)? If so, is there a version (and possibly a PKGBUILD ) of APP-SDK that works with the legacy driver?

Vi0L0 · 2013-12-05 17:48:54

iirc 2.7 should be fine: http://catalyst.wirephire.com/tarball/amdstream/

Arch Linux

#5926 2013-11-27 19:03:25

Re: The AMD/ATI Bar & Grill

#5927 2013-11-27 19:22:26

Re: The AMD/ATI Bar & Grill

#5928 2013-11-27 19:30:40

Re: The AMD/ATI Bar & Grill

#5929 2013-11-27 19:38:23

Re: The AMD/ATI Bar & Grill

#5930 2013-11-27 19:52:21

Re: The AMD/ATI Bar & Grill

#5931 2013-11-27 20:15:33

Re: The AMD/ATI Bar & Grill

#5932 2013-11-27 21:01:46

Re: The AMD/ATI Bar & Grill

#5933 2013-11-28 10:20:33

Re: The AMD/ATI Bar & Grill

#5934 2013-11-28 11:01:08

Re: The AMD/ATI Bar & Grill

#5935 2013-11-28 11:21:16

Re: The AMD/ATI Bar & Grill

#5936 2013-11-28 14:31:44

Re: The AMD/ATI Bar & Grill

#5937 2013-11-29 06:24:40

Re: The AMD/ATI Bar & Grill

#5938 2013-11-30 10:28:51

Re: The AMD/ATI Bar & Grill

#5939 2013-11-30 11:42:24

Re: The AMD/ATI Bar & Grill

#5940 2013-12-01 04:41:42

Re: The AMD/ATI Bar & Grill

#5941 2013-12-02 04:15:07

Re: The AMD/ATI Bar & Grill

#5942 2013-12-04 20:48:59

Re: The AMD/ATI Bar & Grill

#5943 2013-12-04 21:39:22

Re: The AMD/ATI Bar & Grill

#5944 2013-12-05 04:18:29

Re: The AMD/ATI Bar & Grill

#5945 2013-12-05 05:27:43

Re: The AMD/ATI Bar & Grill

#5946 2013-12-05 07:48:59

Re: The AMD/ATI Bar & Grill

#5947 2013-12-05 11:54:42

Re: The AMD/ATI Bar & Grill

#5948 2013-12-05 14:18:53

Re: The AMD/ATI Bar & Grill

#5949 2013-12-05 17:25:30

Re: The AMD/ATI Bar & Grill

#5950 2013-12-05 17:48:54

Re: The AMD/ATI Bar & Grill

Board footer