The AMD/ATI Bar & Grill

Cdh · 2010-09-27 00:06:20

VIOLO, you may want to use mediainfo to check how this video was encoded.

Because we had the same problem here:
https://bbs.archlinux.org/viewtopic.php … 93#p826493

It's kind of clear that it happens only x264 High@L5.0... You could verify that?

Vi0L0 · 2010-09-27 01:11:43

Cdh wrote:

VIOLO, you may want to use mediainfo to check how this video was encoded.
Because we had the same problem here:
https://bbs.archlinux.org/viewtopic.php … 93#p826493
It's kind of clear that it happens only x264 High@L5.0... You could verify that?

Well... i can confirm that 1080p mp4 trailer from http://durian.blender.org/download/ also got problems.

But my mkv vid got High@L4.0 Format Profile...

Video
ID                               : 1
Format                           : AVC
Format/Info                      : Advanced Video Codec
Format profile                   : High@L4.0
Format settings, CABAC           : Yes
Format settings, ReFrames        : 5 frames
Muxing mode                      : Container profile=Unknown@4.0
Codec ID                         : V_MPEG4/ISO/AVC
Duration                         : 2h 24mn
Bit rate                         : 3 889 Kbps
Width                            : 1 920 pixels
Height                           : 1 080 pixels
Display aspect ratio             : 16:9
Frame rate                       : 23.976 fps
Color space                      : YUV
Chroma subsampling               : 4:2:0
Bit depth                        : 8 bits
Scan type                        : Progressive
Bits/(Pixel*Frame)               : 0.078
Stream size                      : 3.82 GiB (70%)
Writing library                  : x264 core 67 r1162M f7bfcfa

LukynZ · 2010-09-27 09:40:45

I am trying gstreamer-vaapi, but I am unable to run it => https://bbs.archlinux.org/viewtopic.php … 34#p831834

Did anybody try it?

Last edited by LukynZ (2010-09-27 09:41:16)

subzero316 · 2010-09-27 12:40:17

Vi0L0 wrote:

Cdh wrote:
VIOLO, you may want to use mediainfo to check how this video was encoded.
Because we had the same problem here:
https://bbs.archlinux.org/viewtopic.php … 93#p826493
It's kind of clear that it happens only x264 High@L5.0... You could verify that?
Well... i can confirm that 1080p mp4 trailer from http://durian.blender.org/download/ also got problems.

This video works perfect for me with -vo vaapi:gl -va vaapi
My card is Mobilty Radeon 4570
I do get Too many buffered pts error on mkv files I remove it using --nocorrect-pts argument for mplayer.
I compile mplayer-vaapi libva and xvba-video and dont install them from the repos.

LukynZ · 2010-09-27 19:21:38

Well with last .6 kernel upgrade X can't start. I did catalyst_blabla remove, then all, but can't start. Switching back to .5 and everything works. I am using xorg 1.9. Does anybody have same problem?

Revolt · 2010-09-27 22:33:09

I do. From what I remember (don't have my laptop here with me atm) it errored out with a Segmentation Fault (X11 that is) right after loading fglrx

LukynZ · 2010-09-27 22:39:16

Yes. Thats the problem. Entire log is here http://pastebin.com/AWDmVwTT

Vi0L0 · 2010-09-28 01:41:53

Edit: Problem does not exist on i686 systems
Yes, it's looking like module built for 2.6.35.6(-ARCH) isn't working as it should...
I actually don't know to who should i to thank for this... (i guess it's kernel's devs fault ;P)
I tried some tricks but it's really hard to trace bug when there's not even one error or warning during module compilation...
I will try to fight with this later...

For now: Temporary solution when updating kernel26 2.6.35.5 -> 2.6.35.6

Do NOT rebuild fglrx module after kernel's update (so don't use catalyst_build_module at all)

(solution below was written for catalyst-test)
If you already rebuilt fglrx module and booted new kernel and you can see nice black tty login screen - then:

1. as root downgrade kernel:

pacman -U /var/cache/pacman/pkg/kernel26-headers-2.6.35.5-1-x86_64.pkg.tar.xz
pacman -U /var/cache/pacman/pkg/kernel26-2.6.35.5-1-x86_64.pkg.tar.xz

2. as unprivileged user type:
catalyst_build_module 2.6.35-ARCH

3. as root update kernel:
pacman -Su

4. reboot

It should work fine.

Btw. i am now wondering how will catalyst 10.9 auto re-compilation packages work with new kernel (when it will land in [core])... i guess that disabling auto re-compilation before update is the best way atm...

Edit: oh... 2.6.35.6 already is in [core]... and 10.9 fglrx module built on that kernel isn't working, sigh...
best way atm for catalyst-hook or catalyst-daemon users: use catalyst package from [catalyst] repo...
or walk the steps i wrote above, only that before running step 3 - disable auto-recompilation: catalyst_build_module autooff

Last edited by Vi0L0 (2010-09-28 03:35:47)

cyberpatrol · 2010-09-28 04:01:41

Vi0L0 wrote:

Edit: Problem does not exist on i686 systems
Yes, it's looking like module built for 2.6.35.6(-ARCH) isn't working as it should...
I actually don't know to who should i to thank for this... (i guess it's kernel's devs fault ;P)
I tried some tricks but it's really hard to trace bug when there's not even one error or warning during module compilation...
I will try to fight with this later...

Since virtualbox doesn't seem to work with the new kernel, too, I guess it's a bug or a new feature in the kernel.

mattoufoutu · 2010-09-28 09:59:53

Vi0LO wrote:

Edit: Problem does not exist on i686 systems

So you mean it's safe to upgrade to 2.6.35.6 on i686 systems?
(that's how i understand it, but i still prefer to be sure )

transmutated · 2010-09-28 12:48:00

Vi0L0 wrote:

Do NOT rebuild fglrx module after kernel's update (so don't use catalyst_build_module at all)
(solution below was written for catalyst-test)
If you already rebuilt fglrx module and booted new kernel and you can see nice black tty login screen - then:

For what its worth, I'm functioning as I should be with the new catalyst and kernel. I using the daemon instead of the hook. I am currently running a Radeon HD 5850. How can I check my version to be sure I am running 10.9? When I check catalyst, it shows my original version from my first install?

Thanks for the great packages Vi0l0!

LeCrayonVert · 2010-09-28 12:54:20

transmutated wrote:

Vi0L0 wrote:
Do NOT rebuild fglrx module after kernel's update (so don't use catalyst_build_module at all)
(solution below was written for catalyst-test)
If you already rebuilt fglrx module and booted new kernel and you can see nice black tty login screen - then:
For what its worth, I'm functioning as I should be with the new catalyst and kernel. I using the daemon instead of the hook. I am currently running a Radeon HD 5850. How can I check my version to be sure I am running 10.9? When I check catalyst, it shows my original version from my first install?
Thanks for the great packages Vi0l0!

http://wiki.archlinux.org/index.php/ATI … 7s_version

Last edited by LeCrayonVert (2010-09-28 12:56:35)

msx · 2010-09-28 13:00:03

"[...] and you can see nice black tty login screen [...]" xD hey, that's me!! haha

ChemBro · 2010-09-28 13:10:05

mattoufoutu wrote:

Vi0LO wrote:
Edit: Problem does not exist on i686 systems
So you mean it's safe to upgrade to 2.6.35.6 on i686 systems?
(that's how i understand it, but i still prefer to be sure )

It worked on my i686 system.

Vi0L0 · 2010-09-28 15:51:48

cyberpatrol wrote:

Vi0L0 wrote:
Edit: Problem does not exist on i686 systems
Yes, it's looking like module built for 2.6.35.6(-ARCH) isn't working as it should...
I actually don't know to who should i to thank for this... (i guess it's kernel's devs fault ;P)
I tried some tricks but it's really hard to trace bug when there's not even one error or warning during module compilation...
I will try to fight with this later...
Since virtualbox doesn't seem to work with the new kernel, too, I guess it's a bug or a new feature in the kernel.

I've found a bug in our kernel26-headers-2.6.35.6-1-x86_64 package, and sent mail to it's maintainer.
When it will be fixed - catalyst, virtualbox and vmware modules should work fine.

Edit: fglrx module (and vbox modules) built on 2.6.35.6-2 version of kernel26 and kernel26-headers (which are atm in [testing]) works fine.

Last edited by Vi0L0 (2010-09-28 19:45:02)

agapito · 2010-09-29 08:27:31

XSERVER 1.9 ARRIVED TO EXTRA. DO NOT UPGRADE IT OR INSTALL CATALYST-TEST PACKAGE FROM AUR WITH 1.9 SUPPORT.

On my machine 1.9 + catalyst-test works better than 1.8 + catalyst 10.9.

v6ak · 2010-09-29 08:54:22

What is better? Performance? Power managgement? Fixed bugs?

McFlow · 2010-09-29 11:16:38

So, it's me again. Never ending segfaults.
I did a system upgrade today. I saw there is an update to Xorg 1.9 but Catalyst 10.5 had version conflict with it. So I removed it and hoped latest catalyst and Xorg 1.9 will finally work on my system.
Of course i commented eveything related to xorg and clyde out in clyde.conf. The only additional repo I have enabled is multilib.
After this update I rebooted my machine and installed the catalyst-test from AUR and ran the command that it told me i should run. Then I generated a xorg.conf with aticonfig --initial --force and rebooted again. The Xserver did not start up.
This is /var/log/Xorg.0.log

#
[    26.432] (EE) fglrx(0): swlDriScreenInit failed!
#
[    26.432] (WW) fglrx(0): ***********************************************************
#
[    26.432] (WW) fglrx(0): * DRI initialization failed                               *
#
[    26.432] (WW) fglrx(0): * kernel module (fglrx.ko) may be missing or incompatible *
#
[    26.432] (WW) fglrx(0): * 2D and 3D acceleration disabled                         *
#
[    26.432] (WW) fglrx(0): ***********************************************************

This message is new to me

After removing catalyst-test I noticed that fglrx.ko is still present. So I removed it by hand and reinstalled catalyst-test. No change. Looks like I have to go back to xorg-1.8 and catalyst 10.5. That's the latest combination that works for me.

Edit:
Ouch. Xorg18 and catalyst 10.5 don't work. It tells "no screens found". And Xorg17 and Xorg 10.5 gives same error as above. Looks like i would do a fresh install...

Last edited by McFlow (2010-09-29 12:50:59)

Vi0L0 · 2010-09-29 13:45:57

@McFlow: Stop! Let me explain...

atm i cannot test does kernel 2.6.35.7 is working fine (i'm out of my base ) but have you tested [xorg18] repo (read ati catalyst wiki -> "Catalyst's repositories" to get more info) with catalyst 10.5? it was working fine when i was testing it lately.
Also: are you sure that fglrx module of 10.5 was built succesfuly? I mean there's a arch_compat.patch to work with newest 2.6.35 kernels

Edit: this error:
fglrx(0): swlDriScreenInit failed!
occurs when using buggy kernel26-headers 2.6.35.6-1, are you sure you updated your system lately?

Edit: i've just tested catalyst-test with fresh kernel26-headers 2.6.35.7 and it's working fine

Last edited by Vi0L0 (2010-09-29 14:21:12)

McFlow · 2010-09-29 14:25:10

I built the catalyst 10.5 myself by changing the kernel dependency in PKGBUILD... With the ones from the catalyst-10.5 repo it complains that the kernel is too new. So i thought this might be right. Before it worked...

Yes, I seem to use kernel-headers in this version.

[florian@mydell ~]$ sudo clyde -Q kernel26-headers
core/kernel26-headers 2.6.35.6-1

Maybe it's just the headers issue then...
But there is no update offered to me. I have to investigate my clyde.conf

Last edited by McFlow (2010-09-29 14:26:59)

Vi0L0 · 2010-09-29 14:33:04

McFlow wrote:

I built the catalyst 10.5 myself by changing the kernel dependency in PKGBUILD... With the ones from the catalyst-10.5 repo it complains that the kernel is too new. So i thought this might be right. Before it worked...

well... it won't work on newest kernels without arch_compat.patch

McFlow wrote:

Yes, I seem to use kernel-headers in this version.
[florian@mydell ~]$ sudo clyde -Q kernel26-headers
core/kernel26-headers 2.6.35.6-1
Maybe it's just the headers issue then...

surely it was kernel's headers issue, read my post above, #2516

McFlow wrote:

But there is no update offered to me. I have to investigate my clyde.conf

maybe your mirror isn't synced, new packages are already available at:
ftp://ftp.archlinux.org/core/os/x86_64/

Last edited by Vi0L0 (2010-09-29 14:36:03)

McFlow · 2010-09-29 15:34:14

There was a kernel header issue. Yes. I took the mirrorlist from my desktop pc and I got the update of it.
After I updated the Xserver still crashed with segfault signal 11. As I almost expected.
About Xorg18 and Catalyst 15.
I added xorg18 repo and downloaded catalyst 10.5 from here. Then I took the compat patch from the 10.9 package in AUR and edited the PKGBUILD (patch source, md5, patch reference)
It built fine and after installing and reboot Xserver did not start up. xorg.0.log.
I changed to Xorg17 repo and it worked then. For now I have a stable system I can use but still I want to figure out why the new catalyst drivers make Xserver just segfaulting! There was no activity on the bugreport i opened yet.

Thanks for your help.

Vi0L0 · 2010-09-29 22:03:40

[catalyst] repo:
for those who are catching artifacts with catalyst-test + xorg-server 1.9 company
i have just added xorg-server-catalyst-maximize-fix 1.9.0-1 package (backclear patch)
note: i've no idea how it will work with other xorg-server packages (like xorg-server-xdmx) from [extra]

McFlow wrote:

For now I have a stable system I can use but still I want to figure out why the new catalyst drivers make Xserver just segfaulting!

Did you check xf86-video-ati to be sure it's catalyst's fault?

If you are sure, then i guess i can prepare for you catalyst 10.9 package which will work with xserver 1.7, edit: wait, not with 1.7 - but with 1.6

Also: did you check is 10.9 working on xserver 1.7?

I'm pretty sure that your problem is somehow related with udev support introduced in xserv 1.8.

Another question: which catalyst-10.5 repo were you using? this one with xserv 1.7, or this one with xserv 1.8?

Last edited by Vi0L0 (2010-09-29 22:54:35)

DarksideEE7 · 2010-09-29 22:43:37

So what is the consensus on xorg 1.9 vs Catalyst 10.9, 10.10, or else?

Vi0L0 · 2010-09-29 22:52:59

DarksideEE7 wrote:

So what is the consensus on xorg 1.9 vs Catalyst 10.9, 10.10, or else?

10.9 doesn't support xserv 1.9 at all, so either use:
* catalyst-test
or
* 10.9 + [xorg18] repo

i think that 10.10 will support xserv 1.9 when it will be finally released

ps. i versioned catalyst-test as 10.10-xxx because ubuntu 10.10 is it's target

Arch Linux

#2501 2010-09-27 00:06:20

Re: The AMD/ATI Bar & Grill

#2502 2010-09-27 01:11:43

Re: The AMD/ATI Bar & Grill

#2503 2010-09-27 09:40:45

Re: The AMD/ATI Bar & Grill

#2504 2010-09-27 12:40:17

Re: The AMD/ATI Bar & Grill

#2505 2010-09-27 19:21:38

Re: The AMD/ATI Bar & Grill

#2506 2010-09-27 22:33:09

Re: The AMD/ATI Bar & Grill

#2507 2010-09-27 22:39:16

Re: The AMD/ATI Bar & Grill

#2508 2010-09-28 01:41:53

Re: The AMD/ATI Bar & Grill

#2509 2010-09-28 04:01:41

Re: The AMD/ATI Bar & Grill

#2510 2010-09-28 09:59:53

Re: The AMD/ATI Bar & Grill

#2511 2010-09-28 12:48:00

Re: The AMD/ATI Bar & Grill

#2512 2010-09-28 12:54:20

Re: The AMD/ATI Bar & Grill

#2513 2010-09-28 13:00:03

Re: The AMD/ATI Bar & Grill

#2514 2010-09-28 13:10:05

Re: The AMD/ATI Bar & Grill

#2515 2010-09-28 15:51:48

Re: The AMD/ATI Bar & Grill

#2516 2010-09-29 08:27:31

Re: The AMD/ATI Bar & Grill

#2517 2010-09-29 08:54:22

Re: The AMD/ATI Bar & Grill

#2518 2010-09-29 11:16:38

Re: The AMD/ATI Bar & Grill

#2519 2010-09-29 13:45:57

Re: The AMD/ATI Bar & Grill

#2520 2010-09-29 14:25:10

Re: The AMD/ATI Bar & Grill

#2521 2010-09-29 14:33:04

Re: The AMD/ATI Bar & Grill

#2522 2010-09-29 15:34:14

Re: The AMD/ATI Bar & Grill

#2523 2010-09-29 22:03:40

Re: The AMD/ATI Bar & Grill

#2524 2010-09-29 22:43:37

Re: The AMD/ATI Bar & Grill

#2525 2010-09-29 22:52:59

Re: The AMD/ATI Bar & Grill

Board footer