The AMD/ATI Bar & Grill

snack · 2014-01-07 09:32:51

Vi0L0 wrote:

lano1106 wrote:
You know that the end of catalyst support is approaching when its maintainer starts to appreciate the oss drivers :-)
I would greet this day with a smile on my face :-).
I'm sure they do appreciate the oss drivers already, guess that what keeps catalyst alive is (mainly) the support of corporate customers (opengl 4.x and opencl support is important there).

Man, I guess he's talking about you (the Catalyst maintainer for Arch) and not about the AMD devs...

Vi0L0 · 2014-01-07 09:48:35

snack wrote:

Vi0L0 wrote:
lano1106 wrote:
You know that the end of catalyst support is approaching when its maintainer starts to appreciate the oss drivers :-)
I would greet this day with a smile on my face :-).
I'm sure they do appreciate the oss drivers already, guess that what keeps catalyst alive is (mainly) the support of corporate customers (opengl 4.x and opencl support is important there).
Man, I guess he's talking about you (the Catalyst maintainer for Arch) and not about the AMD devs...

Oh, could be I missred, LOL. I think that I will drop catalyst... some day :-)
As for now it's still in use, I still like to maintain it, and I didn't get a single e-mail from someone interested in taking it back from me...

clfarron4 · 2014-01-07 12:07:50

Vi0L0 wrote:

As for now it's still in use, I still like to maintain it, and I didn't get a single e-mail from someone interested in taking it back from me...

Give me a few months to get used to PKGBUILDs and I'll think about it.

Vi0L0 · 2014-01-07 12:25:04

clfarron4 wrote:

Vi0L0 wrote:
As for now it's still in use, I still like to maintain it, and I didn't get a single e-mail from someone interested in taking it back from me...
Give me a few months to get used to PKGBUILDs and I'll think about it.

Sure, take your time. I can also give you some tips and scripts when you will be ready.

hippieben · 2014-01-07 18:43:16

I have the group xorg set to be ignored by pacman, however I just checked for updates this is the output

ben@arch-desktop:~$ sudo pacman -Syu
[sudo] password for ben: 
:: Synchronizing package databases...
 catalyst is up to date
 core                     106.4 KiB   104K/s 00:01 [#####################] 100%
 extra                   1441.8 KiB   255K/s 00:06 [#####################] 100%
 community                  2.0 MiB   315K/s 00:07 [#####################] 100%
 multilib                 112.2 KiB   263K/s 00:00 [#####################] 100%
:: Starting full system upgrade...
warning: xorg-server: ignoring package upgrade (1.14.5-1 => 1.14.5-2)
resolving dependencies...
looking for inter-conflicts...

Packages (10): gst-plugins-base-1.2.2-1  gst-plugins-base-libs-1.2.2-1
               gst-plugins-good-1.2.2-1  gstreamer-1.2.2-1
               kdebase-workspace-4.11.5-1  libanjuta-3.10.2-2
               libxfont-1.4.7-1  mesa-10.0.1-2  openssl-1.0.1.f-1
               xorg-server-common-1.14.5-2

Total Download Size:    26.99 MiB
Total Installed Size:   122.15 MiB
Net Upgrade Size:       -1.76 MiB

:: Proceed with installation? [Y/n]

As you can see pacman is warning me that it will not upgrade xorg-server, yet wants me to install xorg-server-common-1.14.5-2. Is this safe to do or should I add xorg-server-common to the list of packages ignored? Why is it not already included in the xorg group?

lano1106 · 2014-01-07 21:51:33

hippieben wrote:

As you can see pacman is warning me that it will not upgrade xorg-server, yet wants me to install xorg-server-common-1.14.5-2. Is this safe to do or should I add xorg-server-common to the list of packages ignored? Why is it not already included in the xorg group?

lano1106@hpmini ~/dev/alsa-lib.git/src/pcm $ pacman -Ql xorg-server-common
xorg-server-common /usr/
xorg-server-common /usr/lib/
xorg-server-common /usr/lib/xorg/
xorg-server-common /usr/lib/xorg/protocol.txt
xorg-server-common /usr/share/
xorg-server-common /usr/share/licenses/
xorg-server-common /usr/share/licenses/xorg-server-common/
xorg-server-common /usr/share/licenses/xorg-server-common/COPYING
xorg-server-common /usr/share/man/
xorg-server-common /usr/share/man/man1/
xorg-server-common /usr/share/man/man1/Xserver.1.gz
xorg-server-common /var/
xorg-server-common /var/lib/
xorg-server-common /var/lib/xkb/
xorg-server-common /var/lib/xkb/README.compiled

IOW, it only contains documentation. xorg-server-common is the less important package of the bunch. I see no problem if you want to get the official one..

lano1106 · 2014-01-07 22:09:09

Vi0L0 wrote:

I started using threadirqs as soon as it appeared. Back in the days I was using q6600 + p35 and it gave me some nice performance boost - DE was more responsive when transferring data to the usb mass storage.
Then, after bying i7, I simply moved the arch disk from one mb to another, and was still using threadirqs, never whining about systems performance (compiling on all cores in the ramdisk while watching hd movie while copying data, all on vanilla kernel - man, i just love linux ;-P). I never actually checked what's the performance impact of threadirqs (or is it actually needed on i7) but I kept in mind that in the old days threadirqs was good for me.
And man, what was my surprise when I saw that 3.13 rc5 is still freezing X whenever i'm trying to use usb 2.0 mouse...
Anyhow, now i think that doing some benchmarks of threadirqs is a good idea ;-), I will surely do this in a free time and give you a feedback.

I think that back at that time when threadsirq got introduced, irqs were processed on single core insteads of being load balanced among all available cores.

That probably gave a small latency advantage to defer their work as fast as possible to a kernel thread as the irq were kinda serialized.

This is probably less true today.

cat /proc/interrupts can give some insights on irq dispatching.

this is a bit offtopic but please share the results if you have time to benchmark.

joanmanel · 2014-01-08 10:51:19

Hi,

I configured my ATI graphic card more than a year ago when I got my laptop.

It has 2 graphic cards, Radeon HD 7730M and Intel Corporation 3rd Gen Core processor Graphics.

I had a hard time, and spent something like 2 weeks just to get the system working, with the Intel card.

I have installed the aur/catalyst-total-pxp, but it only works for the intel option, when i try to swtich to the AMD, it fails.

I also have pacman.conf set to install and not upgrade more than xorg112.

After a year, I am wondering if anything has changed, if my graphic card is supported, and if I can update my system to use the card.

I cannot "try and see" because I need my computer, and I cannot be "off" two weeks like last time.

ghaith · 2014-01-08 19:52:01

joanmanel wrote:

Hi,
I configured my ATI graphic card more than a year ago when I got my laptop.
It has 2 graphic cards, Radeon HD 7730M and Intel Corporation 3rd Gen Core processor Graphics.
I had a hard time, and spent something like 2 weeks just to get the system working, with the Intel card.
I have installed the aur/catalyst-total-pxp, but it only works for the intel option, when i try to swtich to the AMD, it fails.
I also have pacman.conf set to install and not upgrade more than xorg112.
After a year, I am wondering if anything has changed, if my graphic card is supported, and if I can update my system to use the card.
I cannot "try and see" because I need my computer, and I cannot be "off" two weeks like last time.

I've had mixed results with hybrid graphics and linux, Last arch install, i settled on catalyst-test-pxp with xorg 1.14 until 1.15 came out, now i switched back to intel graphics only with vga switcheroo to turn off the radeon card.
Anyway regarding the installation, I'm assuming you followed the wiki to setup catalyst, if so note that the KMS limitation mentioned there doesn't - as far as I have noticed - apply to pxp, so if you disabled KMS enable it again, I didn't disable it as I remember running into issues when i first disabled it, and ofcourse make sure your radeon module is blacklisted and not in the MODULES section of /etc/mkinitcpio.conf (You would have it there if you followed the wiki article for hybrid graphics)

Lockheed · 2014-01-08 21:49:01

Vi0L0 wrote:

Lately because of the problems with 3.13 kernel' threadirqs and the fact that I had to switch to [testing] I decided to check out oss radeon from lorheavy's mesa-git repo... and wow! How amazed I am!
3.13 rc7 and git drivers are working so beautiful on my hd 7850. All Valve's Steam Games are working fantastic (well I had to disable AA in Dota 2, but who cares ;P). It was such a bad experience when using oss radeon on 3.12 kernel, even kwin was lagging, and now it's jus "wow wow wee waa!". "Go go AMD OSS Rangers" I would say
Edit: notice that I wrote about Valve's Games, Metro LL for example don't even dare to launch - just a FYI
Edit2: darn, (even) chromeexperiment' webgl examples are working well...

That is great news.
However, performance is not the only issue. For laptop users, power consumption is the key, and last time I checked, open source ati drivers were nowhere near catalysts.
So until OSS power management in is on pair with Catalysts, mobile users like myself are forced to stick to Catalysts.

snack · 2014-01-09 07:58:37

Lockheed wrote:

However, performance is not the only issue. For laptop users, power consumption is the key, and last time I checked, open source ati drivers were nowhere near catalysts.
So until OSS power management in is on pair with Catalysts, mobile users like myself are forced to stick to Catalysts.

Now power management on OSS drivers is almost good as in Catalyst, if not equal. It is surely worth a try, but be aware that if you own an old RV6xx card you might run into this issue

Last edited by snack (2014-01-09 07:59:08)

Alaska23 · 2014-01-09 23:14:24

Vi0L0 wrote:

Lately because of the problems with 3.13 kernel' threadirqs and the fact that I had to switch to [testing] I decided to check out oss radeon from lorheavy's mesa-git repo... and wow! How amazed I am!
3.13 rc7 and git drivers are working so beautiful on my hd 7850. All Valve's Steam Games are working fantastic (well I had to disable AA in Dota 2, but who cares ;P). It was such a bad experience when using oss radeon on 3.12 kernel, even kwin was lagging, and now it's jus "wow wow wee waa!". "Go go AMD OSS Rangers" I would say
Edit: notice that I wrote about Valve's Games, Metro LL for example don't even dare to launch - just a FYI
Edit2: darn, (even) chromeexperiment' webgl examples are working well...

nice to hear that.
I've got the "msi HD7850 TwinFrozr OC" and tried the OSS-drivers a few weeks back with 3.12. It was already very promising and the idle-power consumption was just a few Watts above the catalyst-driver. What i counldn't get fixed is the fanspeed-adaption by temperature, same problem as this guy has:
http://phoronix.com/forums/showthread.p … ith-radeon
The standard fanspeed is very annoying and causes headaches after a few minutes. It didn't chance even if the sensors showed a temperature just around 30 degrees.
I'll try it again when 3.13 final is released.

Vi0L0 · 2014-01-10 23:46:12

catalyst 13.11 beta 9.95, which I called catalyst 13.111 cuz it's easier to maintain...

Guys, there's this new driver which firstly appeared on SteamOS, and I must to say that for me it's the best catalyst since many many months, maybe ever.
It looks like AMD has finally manage to get rid of that nasty "overlay" bug, which appeared whenever some opengl/vaapi/other_specific fullscreen application was trying to draw frames on the (kde) composited desktop.
Bad things then happened, ie:
- blue screen when playing html5 videos;
- OSD glitches when watching hd movie via smplayers vaapi;
- OSD glitches when using gwenview and trying to click on the thumbnail;
- flickering whenever volume bar appeared over flash video;
- etc etc.
I was loosing hope on this but it's finally working as it should!
The performance of Steam games also looks better.

If you are using catalyst-stable I recommand to switch to catalyst-test@AUR or [catalyst] repo, just give it a shot. If it's not working well for you, you can always downgrade to catalyst-stable.
Just a tip:
At first I wasn't able to use opengl mode in KDEs kwin (sic!). I had to remove /etc/ati/amdpcsdb file, this solved the problem.

Last edited by Vi0L0 (2014-01-10 23:57:28)

owend · 2014-01-11 02:21:38

I installed catalyst-hook earlier this week on a dual boot Mac. I confirmed I had the new drivers working (glxinfo reported OpenGL 4 support), but when I rebooted I had no video. Is it possible the catalyst installation process did something to make the card invisible to the Mac's EFI (like a PC version of the card apparently would be)? Or is it maybe just a coincidence that the card failed at this time?

(I'm relatively new to Arch. I apologize if this is the wrong place to post a question like this.)

shazeal · 2014-01-11 19:02:52

Vi0L0 wrote:

Guys, there's this new driver which firstly appeared on SteamOS, and I must to say that for me it's the best catalyst since many many months, maybe ever.

Have to agree this fixed all the issues I was having with KDE, which weren't many or major in form to begin with. Main annoyance was general corruption (chequered pattern) which affected every "effect" in KDE. is now gone. In general it does seem a lot smoother, no flickers and such either on effect transitions.

Last edited by shazeal (2014-01-11 19:04:39)

scjet · 2014-01-11 20:31:03

I always use "catalyst-test@AUR".

solar · 2014-01-12 08:16:44

Well Violo.. czescz maly ,p

I don't know... I did the total package from aur, and now I get th eold pxp error again. I initially got this by *using* pxp, and all was golden (on a desktop, with igpu, but the point is, less hassle as no need ot remove all the intel stuff).. however... amdcccle seemed to be working up.. so I reinstalled, and now only used the total package, but still get the pxp error.. bla

not to mention, amdcccle doesn't ever remember anything

EDIT: Ok, seems to be a regression.

Workaround which worked here, was A. use the pxp version (even on desktop, as I have an onboard gpu).

then crash:
grep EE /var/log/Xorg.0.log

I got this: (EE) AIGLX error: failed to open /usr/X11R6/lib64/modules/dri/fglrx_dri.so, error[/usr/X11R6/lib64/modules/dri/fglrx_dri.so: cannot open shared object file: No such file or directory]

sudo mkdir /usr/X11R6
sudo ln -s /usr/lib/xorg /usr/X11R6/lib64

Again, this is not exactly a nice work-around but maybe due to some updates... anyway, I am not sure how it is, without the pxp , just pure total as it shoudl be on a desktop.. but it might help those who get that same error above.

ps. I believe the aiglx is btw in one of the *intel* packages, but I cba to reinstall without pxp just to confirm.

Last edited by solar (2014-01-12 09:25:24)

snack · 2014-01-12 11:31:20

I just updated my stock kernel to 3.12.7-1 and I got the usual, annoying error messages from catalyst-hook during the pacman run:

(5/7) upgrading linux                                                                                        [################################################################] 100%
>>> Updating module dependencies. Please wait ...
>>> Generating initial ramdisk, using mkinitcpio.  Please wait...
==> Building image from preset: /etc/mkinitcpio.d/linux.preset: 'default'
  -> -k /boot/vmlinuz-linux -c /etc/mkinitcpio.conf -g /boot/initramfs-linux.img
==> Starting build: 3.12.7-1-ARCH
  -> Running build hook: [base]
  -> Running build hook: [udev]
  -> Running build hook: [autodetect]
  -> Running build hook: [block]
  -> Running build hook: [fglrx]
Building fglrx module for 3.12.7-1-ARCH kernel ...
Failed!!! Check out log: /var/log/catalyst-install.log
- /usr/lib/modules/3.12.6-1-ARCH looks like unused, maybe remove it manualy?
==> Generating module dependencies
==> Creating gzip initcpio image: /boot/initramfs-linux.img
==> Image generation successful
==> Building image from preset: /etc/mkinitcpio.d/linux.preset: 'fallback'
  -> -k /boot/vmlinuz-linux -c /etc/mkinitcpio.conf -g /boot/initramfs-linux-fallback.img -S autodetect
==> Starting build: 3.12.7-1-ARCH
  -> Running build hook: [base]
  -> Running build hook: [udev]
  -> Running build hook: [block]
==> WARNING: Possibly missing firmware for module: aic94xx
==> WARNING: Possibly missing firmware for module: smsmdtv
  -> Running build hook: [fglrx]
Building fglrx module for 3.12.7-1-ARCH kernel ...
Failed!!! Check out log: /var/log/catalyst-install.log
- /usr/lib/modules/3.12.6-1-ARCH looks like unused, maybe remove it manualy?
==> Generating module dependencies
==> Creating gzip initcpio image: /boot/initramfs-linux-fallback.img
==> Image generation successful
(6/7) upgrading linux-headers

I always get these errors, the first is due to the fact that the new kernel headers are still not present at the moment of the compilation of fglrx. The second one suggest to manually remove the module folder for the old kernel version, even if AFAIK it is done automatically in a later stage (since at the end of the pacman run that folder is no more present).
Vi0l0, is it possible to disable automatic compilation of fglrx for catalyst-hook, given the facts that it always fail and that it is anyway compiled during shutdown? The warning about manually removing the old modules folder is not annoying like the white-on-blue compilation error, but it is another possible improvements of the PKGBUILD. Thanks.

episyron · 2014-01-12 12:03:34

Yes indeed catalyst-test 13.111-1 works better in full screen video respect. But this one introduces new bug: sometimes when opening chromium with multiple windows, the whole screen gets scrambled with flashing triangles. Running KDE with OpenGL 3.1 compositing on Radeon 5770.

solar · 2014-01-12 12:15:23

I cna also confirm my prposed dirty temp hack only worked till reboot... then troubles in paradise.. no X errors though sigh.

I will add, that this is very possibly cinnamon breakage and not the driver on this end... I guess it is just me with my luck, since I decided to do an install from fresh for first time in about a year, and it seems I just had to choose the wrong moment ^^

Having my main install still it works fine, although I haven't checked if I have upgraded fglrx.

Last edited by solar (2014-01-12 12:28:16)

Vi0L0 · 2014-01-12 12:37:06

@solar:
I am aware of

(EE) AIGLX error: failed to open /usr/X11R6/lib64/modules/dri/fglrx_dri.so, error[/usr/X11R6/lib64/modules/dri/fglrx_dri.so: cannot open shared object file: No such file or directory]

but it never had any negative impact on how X or fglrx works. It's just that fglrx got this path hardcoded, so it claims about it. But it should find the lib either way and just work. I could probably find where the hardcode is made and change it with a hex patch, I just never saw any particular reason of doing this. So does it have negative impact on your system (except from generating the error to the log ;P )?
edit: ahh, could be some kind of unluck...

@snack:
you can turn off the hook with:

catalyst_build_module hookrm

but yes, it could be a good idea to leave mkinitcpio untached when installing catalyst-hook and leave user decide if he needs it or not. It should be done with the next catalyst-hook update (although upgrading will not disable it, so you would run mentioned command on your own).

@episyron:
I didn't experience it. Maybe you would be able to get it fixed by manipulating chrome://flags/ ?
BTW for the first time here chromium-dev is working fine and fast without enabling "Override software rendering list" option. All graphics features (chrome://gpu/) were correctly detected.

solar · 2014-01-12 12:40:42

Ok thanks. Violo... ye... well, it got X working... but true.. I haven't seen this error since early 2012 I think (that's why I mentioned regression).

Either way, I am compiling compiz atm to give the old workhorse a try

I use openbox mainly on my box, but had cinnamon-git which I sometimes used... on the fresh install now, I just used the stable in community.. not sure if something is broken atm.

(I don't use testing repos).

Anyway, I guess we will figure it out soon.

Weekend good?

EDIT: Compiz works fine, so does cairo-dock.. etc.. so ye... it is the cinnamon packages which currenlty breaks.

So just to confirm... I was just a bit unlucky for the one time I decided to do a fresh install ,)

Last edited by solar (2014-01-12 12:55:14)

clfarron4 · 2014-01-12 17:37:35

snack wrote:

I just updated my stock kernel to 3.12.7-1 and I got the usual, annoying error messages from catalyst-hook during the pacman run:
(5/7) upgrading linux
[snip]
(6/7) upgrading linux-headers

The reason you get the errors when you upgrade the kernel is because of this, that is the kernel is installed before the corresponding headers are installed. You can blame the pacman developers for removing the SyncFirst option from pacman (so that the headers could be done first), and a certain Allan McCrae has told me on numerous occasions that it will not be re-instated.

snack wrote:

Vi0l0, is it possible to disable automatic compilation of fglrx for catalyst-hook,

You need to tell mkinitcpio not to compile the module automagically, but I don't that is possible due to the way the modules system is managed.

snack wrote:

given the facts that it always fail and that it is anyway compiled during shutdown?

Provided people that use the package enable the service, it'll be compiled (I've seen many people forget to enable the service).

@solar: I'm having no problems with Cinnamon here with Catalyst-test.

Last edited by clfarron4 (2014-01-12 17:38:16)

snack · 2014-01-12 17:47:10

clfarron4 wrote:

snack wrote:
I just updated my stock kernel to 3.12.7-1 and I got the usual, annoying error messages from catalyst-hook during the pacman run:
(5/7) upgrading linux
[snip]
(6/7) upgrading linux-headers
The reason you get the errors when you upgrade the kernel is because of this, that is the kernel is installed before the corresponding headers are installed. You can blame the pacman developers for removing the SyncFirst option from pacman (so that the headers could be done first), and a certain Allan McCrae has told me on numerous occasions that it will not be re-instated.
snack wrote:
Vi0l0, is it possible to disable automatic compilation of fglrx for catalyst-hook,
You need to tell mkinitcpio not to compile the module automagically, but I don't that is possible due to the way the modules system is managed.
snack wrote:
given the facts that it always fail and that it is anyway compiled during shutdown?
Provided people that use the package enable the service, it'll be compiled (I've seen many people forget to enable the service).
@solar: I'm having no problems with Cinnamon here with Catalyst-test.

I know about SyncFirst, and I think that if a skilled developer like Allan decided to remove it then there must be valid reasons for doing so. I don't know the internals of pacman neither the possible breakages that SyncFirst may introduce, so I trust the decisions of the pacman developers.
About disabling automatic compilation: it is possible, see the previous post by Vi0l0.
About not enabling the service: I don't understand your point. Since compilation during the execution of mkinitcpio will always fail, then keeping it active won't save people who forgot to enable the compile-on-shutdown service...

gee · 2014-01-12 22:04:20

Hey all,
since some updates either Thursday or Friday, I cannot use OpenCL anymore
my system just sees the cpu now...
I tried rolling back everything that changed these days, but same results... and on 2 machines.
Anyone else having the same situation?

Arch Linux

#6076 2014-01-07 09:32:51

Re: The AMD/ATI Bar & Grill

#6077 2014-01-07 09:48:35

Re: The AMD/ATI Bar & Grill

#6078 2014-01-07 12:07:50

Re: The AMD/ATI Bar & Grill

#6079 2014-01-07 12:25:04

Re: The AMD/ATI Bar & Grill

#6080 2014-01-07 18:43:16

Re: The AMD/ATI Bar & Grill

#6081 2014-01-07 21:51:33

Re: The AMD/ATI Bar & Grill

#6082 2014-01-07 22:09:09

Re: The AMD/ATI Bar & Grill

#6083 2014-01-08 10:51:19

Re: The AMD/ATI Bar & Grill

#6084 2014-01-08 19:52:01

Re: The AMD/ATI Bar & Grill

#6085 2014-01-08 21:49:01

Re: The AMD/ATI Bar & Grill

#6086 2014-01-09 07:58:37

Re: The AMD/ATI Bar & Grill

#6087 2014-01-09 23:14:24

Re: The AMD/ATI Bar & Grill

#6088 2014-01-10 23:46:12

Re: The AMD/ATI Bar & Grill

#6089 2014-01-11 02:21:38

Re: The AMD/ATI Bar & Grill

#6090 2014-01-11 19:02:52

Re: The AMD/ATI Bar & Grill

#6091 2014-01-11 20:31:03

Re: The AMD/ATI Bar & Grill

#6092 2014-01-12 08:16:44

Re: The AMD/ATI Bar & Grill

#6093 2014-01-12 11:31:20

Re: The AMD/ATI Bar & Grill

#6094 2014-01-12 12:03:34

Re: The AMD/ATI Bar & Grill

#6095 2014-01-12 12:15:23

Re: The AMD/ATI Bar & Grill

#6096 2014-01-12 12:37:06

Re: The AMD/ATI Bar & Grill

#6097 2014-01-12 12:40:42

Re: The AMD/ATI Bar & Grill

#6098 2014-01-12 17:37:35

Re: The AMD/ATI Bar & Grill

#6099 2014-01-12 17:47:10

Re: The AMD/ATI Bar & Grill

#6100 2014-01-12 22:04:20

Re: The AMD/ATI Bar & Grill

Board footer