Discussion: the ati/radeon open source drivers & radeon repository

Jojonintendo · 2015-12-23 13:55:18

@Lone_Wolf: Following your indications I have now installed: mesa-git, llvm-svn, linux-firmware-git, lib32-llvm-svn. But I can't build lib32-mesa-git, all I get is this:

/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXvMC.so when searching for -lXvMC
/usr/bin/ld: skipping incompatible /usr/lib/libXvMC.so when searching for -lXvMC
/usr/bin/ld: cannot find -lXvMC
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXv.so when searching for -lXv
/usr/bin/ld: skipping incompatible /usr/lib/libXv.so when searching for -lXv
/usr/bin/ld: cannot find -lXv
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXvMCW.so when searching for -lXvMCW
/usr/bin/ld: skipping incompatible /usr/lib/libXvMCW.so when searching for -lXvMCW
/usr/bin/ld: cannot find -lXvMCW
collect2: error: ld returned 1 exit status
Makefile:738 : la recette pour la cible « tests/xvmc_bench » a échouée
make[3]: *** [tests/xvmc_bench] Erreur 1
make[3]: *** Attente des tâches non terminées....
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXvMC.so when searching for -lXvMC
/usr/bin/ld: skipping incompatible /usr/lib/libXvMC.so when searching for -lXvMC
/usr/bin/ld: cannot find -lXvMC
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXv.so when searching for -lXv
/usr/bin/ld: skipping incompatible /usr/lib/libXv.so when searching for -lXv
/usr/bin/ld: cannot find -lXv
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXvMCW.so when searching for -lXvMCW
/usr/bin/ld: skipping incompatible /usr/lib/libXvMCW.so when searching for -lXvMCW
/usr/bin/ld: cannot find -lXvMCW
collect2: error: ld returned 1 exit status
Makefile:726 : la recette pour la cible « tests/test_subpicture » a échouée
make[3]: *** [tests/test_subpicture] Erreur 1
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXvMC.so when searching for -lXvMC
/usr/bin/ld: skipping incompatible /usr/lib/libXvMC.so when searching for -lXvMC
/usr/bin/ld: cannot find -lXvMC
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXv.so when searching for -lXv
/usr/bin/ld: skipping incompatible /usr/lib/libXv.so when searching for -lXv
/usr/bin/ld: cannot find -lXv
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXvMCW.so when searching for -lXvMCW
/usr/bin/ld: skipping incompatible /usr/lib/libXvMCW.so when searching for -lXvMCW
/usr/bin/ld: cannot find -lXvMCW
collect2: error: ld returned 1 exit status
Makefile:714 : la recette pour la cible « tests/test_context » a échouée
make[3]: *** [tests/test_context] Erreur 1
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXvMC.so when searching for -lXvMC
/usr/bin/ld: skipping incompatible /usr/lib/libXvMC.so when searching for -lXvMC
/usr/bin/ld: cannot find -lXvMC
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXv.so when searching for -lXv
/usr/bin/ld: skipping incompatible /usr/lib/libXv.so when searching for -lXv
/usr/bin/ld: cannot find -lXv
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXvMCW.so when searching for -lXvMCW
/usr/bin/ld: skipping incompatible /usr/lib/libXvMCW.so when searching for -lXvMCW
/usr/bin/ld: cannot find -lXvMCW
collect2: error: ld returned 1 exit status
Makefile:732 : la recette pour la cible « tests/test_surface » a échouée
make[3]: *** [tests/test_surface] Erreur 1
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXvMC.so when searching for -lXvMC
/usr/bin/ld: skipping incompatible /usr/lib/libXvMC.so when searching for -lXvMC
/usr/bin/ld: cannot find -lXvMC
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXv.so when searching for -lXv
/usr/bin/ld: skipping incompatible /usr/lib/libXv.so when searching for -lXv
/usr/bin/ld: cannot find -lXv
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXvMCW.so when searching for -lXvMCW
/usr/bin/ld: skipping incompatible /usr/lib/libXvMCW.so when searching for -lXvMCW
/usr/bin/ld: cannot find -lXvMCW
collect2: error: ld returned 1 exit status
Makefile:708 : la recette pour la cible « tests/test_blocks » a échouée
make[3]: *** [tests/test_blocks] Erreur 1
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXvMC.so when searching for -lXvMC
/usr/bin/ld: skipping incompatible /usr/lib/libXvMC.so when searching for -lXvMC
/usr/bin/ld: cannot find -lXvMC
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXv.so when searching for -lXv
/usr/bin/ld: skipping incompatible /usr/lib/libXv.so when searching for -lXv
/usr/bin/ld: cannot find -lXv
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-unknown-linux-gnu/5.3.0/../../../libXvMCW.so when searching for -lXvMCW
/usr/bin/ld: skipping incompatible /usr/lib/libXvMCW.so when searching for -lXvMCW
/usr/bin/ld: cannot find -lXvMCW
collect2: error: ld returned 1 exit status
Makefile:720 : la recette pour la cible « tests/test_rendering » a échouée
make[3]: *** [tests/test_rendering] Erreur 1
make[3] : on quitte le répertoire « /home/jojonintendo/Documents/Build/lib32-mesa-git/src/mesa/src/gallium/state_trackers/xvmc »
Makefile:592 : la recette pour la cible « all-recursive » a échouée
make[2]: *** [all-recursive] Erreur 1
make[2] : on quitte le répertoire « /home/jojonintendo/Documents/Build/lib32-mesa-git/src/mesa/src/gallium »
Makefile:685 : la recette pour la cible « all-recursive » a échouée
make[1]: *** [all-recursive] Erreur 1
make[1] : on quitte le répertoire « /home/jojonintendo/Documents/Build/lib32-mesa-git/src/mesa/src »
Makefile:630 : la recette pour la cible « all-recursive » a échouée
make: *** [all-recursive] Erreur 1
==> ERREUR : Une erreur s’est produite dans build().
    Abandon...
[me@Arch lib32-mesa-git]$

Sorry it's in french, I'm not sure how to set it in english.

N.B: I use "pacaur" to check if there are any updates in AUR, but I prefer to use makepkg to go one package at a time.

Edit: ok, forget about it, I should have checked the dependencies first. I installed lib32-libtxvmc and it went good. Sorry, this was an easy one

Edit 2: I still get VM errors and crash with games and videos. I have one question though, about the linux-firmware-git package. It's maintainer says in the comments that it does the same thing as the one in [core]. How can I modify PKGBUILD to actually fetch latest firmware files? Some advice would be very welcome

Last edited by Jojonintendo (2015-12-23 14:11:49)

Lone_Wolf · 2015-12-23 14:42:18

@ Folatt
The HD 8400 is supported by xf86-video-ati .
I suggest you start a separate thread (kernel & hardware or Newbie corner board) , in your opening post include journalctl -b , lspci -k and Xorg.o.log .

@ sawyerbergeron
your problems are looking more and more as a problem with KDE/plasma .
You could try creating a separate thread about it in "Applications & Desktop Environments" , titled something like "no effects with Plasma / Xrender" .
look for an .xsession-errors file somewhere in your home-folder and post that one.

@ jojonintendo :
The aur linux-firmware-git package does build latest version while the [core] linux-firmware builds a specific commit.
Are you using vmwgfx driver ?

sawyerbergeron · 2015-12-23 15:00:08

Lone_Wolf wrote:

@ sawyerbergeron
your problems are looking more and more as a problem with KDE/plasma .
You could try creating a separate thread about it in "Applications & Desktop Environments" , titled something like "no effects with Plasma / Xrender" .
look for an .xsession-errors file somewhere in your home-folder and post that one.

But I get the same exact problem with Gnome 3. Is there any way I can try an old(er) version of the graphics stack and see if it just has to do with a recent version? It works on Ubuntu and such, which I'm pretty sure has older packages in their repos, so would it be worth a shot to try that?
Sorry to bother you with this

Last edited by sawyerbergeron (2015-12-23 15:00:44)

Jojonintendo · 2015-12-23 17:07:01

@Lone_Wolf: I don't even know what vmwgfx is actually . I guess it's related to virtualization, but I didn't install (at least not intentionally) anything vm-related. Plus pacman can't find anything related.

Lone_Wolf · 2015-12-24 12:37:27

I have upload libdrm-git and lib32-libdrm-git to AUR, so latest mesa-git supports all cards again.

@ jojonintendo

I've been advised here to use latest llvm and mesa-git to avoid VM issues, but I still get these errors.

That remark made me guess your issues could be related to vmware / vmwgfx driver.
I do suggest you create a separate thread about the VM issues.
include which virtualization sw you use (qemu, virtualbox, xen etc), info about your host system, what OS you're running in the VM and how you start the VM.

@ sawyerbergeron
Ok,so your problems are with compositing window managers used by gnome (mutter) and KDE (kwin ) ?

downgrading is possible , check https://wiki.archlinux.org/index.php/Do … g_packages .
keep in mind that you may have to downgrade many things.
You should start with determining ubuntu's libdrm, mesa, xorg and kernel versions .

Jojonintendo · 2015-12-24 12:52:11

@Lone_Wolf: Thanks for your help, but I'm not sure I explained it correctly. English is not my native language, so forgive me if I don't express myself the best way . I actually don't use virtualization software, I only have an Arch Linux install on full SSD. "Pacman -Qs" finds nothing about VM and only shows llvm packages. The only thing that comes to my mind that looks like vm software is "wine-staging", that is actually installed on my system. Still, if I understand correctly, WINE doesn't virtualize anything and shouldn't cause me such problems.

Thanks again for your time and help.

Lone_Wolf · 2015-12-24 14:17:06

jojonintendo, I read back on your previous posts and you have problems starting games, and tried latest firmware / mesa / llvm hoping that would solve them ?

If that's correct, then i think you should start a new thread in multimedia and games where you describe the problems as they may be specific to those games (or the framework they use,like steam)

Jojonintendo · 2015-12-24 14:29:57

Actually even videos often lead to same crashes, or other games line unvanquished or wolfenstein: enemy territory. It doesn't seem to be related to specific games or video formats, etc. Youtube leads to same result when playing 1080p 60fps. One common line in dmesg on every crash is "dpm fail" related.

So yes, I'm testing latest firmware, mesa, etc. to see if it gets better. It seems that Grenada gpus don't use exactly the same hardware as Hawaii ones, but lspci recognizes mine as Hawaii.

I'm going to build latest packages from time to time and report back when/if I notice some improvement, for knowledge's sake if anyone is also using Grenada card

Thanks again for your time and help.

sawyerbergeron · 2015-12-24 15:30:48

@Lone_Wolf OK, thanks for the help. I'll start there and see if it works :fingers-crossed:

Just encountered the issue on a fresh install of Ubuntu, so it is definitely an upstream bug. I've submitted a bug report and will use an older Ubuntu install until the issues are resolved, then reinstall Arch.

Last edited by sawyerbergeron (2015-12-29 17:01:27)

WFV · 2016-01-02 23:02:56

I've tried changing all the dram settings in bios to resolve a "no agp aperture, enable iommu" boot error messages to no avail - there is no "iommu" feature in the Asus M5A88-M that I can find with latest bios rev, the dram settings seemed to be closest. System boots without any apparent issues regardless of the message (journalctl -b):

kernel: AGP: Checking aperture...
kernel: AGP: No AGP bridge found
kernel: AGP: Node 0: aperture [bus addr 0x00000000-0x01ffffff] (32MB)
kernel: AGP: Your BIOS doesn't leave an aperture memory hole
kernel: AGP: Please enable the IOMMU option in the BIOS setup
kernel: AGP: This costs you 64MB of RAM
kernel: AGP: Mapping aperture over RAM [mem 0xc4000000-0xc7ffffff] (65536KB)

also tried adding "iommu=noagp" "iommu=noaperture" and "iommu=soft" at separate times to /etc/grub.d/40_custom and update grub and reboot but error remains, probably can be ignored? I set all the dram settings back to their bios defaults and removed the custom grub line.
Is this just an AMD problem?

Last edited by WFV (2016-01-09 23:02:12)

Lone_Wolf · 2016-01-03 16:52:27

It seems ASUS uses another name for AMD-V hw virtualization support and calls it secure virtual machine mode [1] .

Asus M5A88-M user manual [2] has this option under advanced menu - CPU configuration.

I don't know how much difference it will make for your system, but there have been plenty reports of problems about running x86_64 OSes (especially linuxes) that were solved by enabling IOMMU / virtualization support.

[1] https://vardars.wordpress.com/2012/02/1 … a88-v-evo/
[2] https://vardars.wordpress.com/2012/02/1 … a88-v-evo/

WFV · 2016-01-03 18:28:53

Thanks Lone Wolf. I have svm enabled in the bios and an install of Win7pro64bit with Office13 64bit run without issue although I notice some boot messages indicating conflicts of 32 vs 64bit stuff (incl vbox), I'm not having any problems running them and haven't noticed any performance issues either. Also I find in current journalctl a ways after the initial no agp found messages:

kernel: Unpacking initramfs...
kernel: Freeing initrd memory: 3500K (ffff88003791a000 - ffff880037c85000)
kernel: PCI-DMA: Disabling AGP.
kernel: PCI-DMA: aperture base @ c4000000 size 65536 KB
kernel: PCI-DMA: using GART IOMMU.
kernel: PCI-DMA: Reserving 64MB of IOMMU area in the AGP aperture
kernel: perf: AMD NB counters detected

that the system more less ignores the no AGP found and uses the GART IOMMU, so I'm thinking there is no real problem here, seems the kernel works around the AMD shortcomings.

Last edited by WFV (2016-01-04 06:09:17)

lordheavy · 2016-01-25 19:55:41

I've built linux kernel 4.5rc1 with amdgpu powerplay. You can find it here: http://pkgbuild.com/~lcarlier/test/

You'll need also to add amdgpu.powerplay=1 in the commandline in grub/syslinux/...

See also http://www.phoronix.com/scan.php?page=n … GPU-Kernel

[    4.316886] amdgpu: powerplay initialized
[    4.317431] [drm] AMDGPU Display Connectors

haagch · 2016-02-07 12:21:00

lordheavy wrote:

haagch wrote:
@lordheavy
For the time being, can you compile llvm with -mstackrealign like this?
-DCMAKE_C_FLAGS:STRING="-m32 -mstackrealign" \
-DCMAKE_CXX_FLAGS:STRING="-m32 -mstackrealign" \
https://bugs.freedesktop.org/show_bug.cgi?id=93425
Sure, i can do that

With this commit -mstackrealign may not be needed anymore (not tested by me) https://github.com/iXit/Mesa-3D/commit/ … 6496fbb4bd
But dev says it should: https://github.com/iXit/Mesa-3D/issues/ … -180998393

lordheavy · 2016-02-07 20:13:45

haagch wrote:

With this commit -mstackrealign may not be needed anymore (not tested by me) https://github.com/iXit/Mesa-3D/commit/ … 6496fbb4bd
But dev says it should: https://github.com/iXit/Mesa-3D/issues/ … -180998393

I've pushed lib32-llvm-svn without -mstackrealign and lib32-mesa-git with the nine patch.

ChemBro · 2016-02-09 13:42:19

Is ArrayofArrays now supported by the radeonsi driver? I tested it this morning and I can start Shadow of Mordor without the override launch option, but about 1/3 of the textures are missing in the game (and rain is still missing).

lordheavy · 2016-02-12 10:37:38

ChemBro wrote:

Is ArrayofArrays now supported by the radeonsi driver? I tested it this morning and I can start Shadow of Mordor without the override launch option, but about 1/3 of the textures are missing in the game (and rain is still missing).

It still needs force_glsl_extension_warn=true

ChemBro · 2016-02-12 16:21:20

So AoA still does not work for radeonsi, even though it is marked as supported? Bummer. Thought I don't need this override thing anymore.

gee · 2016-03-09 08:30:22

@lordheavy: I'm running into an issue with llvm-svn from your repos, and I think it's related to this bug:
https://bugs.archlinux.org/task/46468

My issue is the same as the bug, I am trying to build a package with cmake, and it complains about the absence of LLVMHello.so.

also this line seems not good either:
rm -f "$pkgdir"/usr/lib/{libLLVM,libLTO,LLVMgold,BugpointPasses}.so

BugpointPasses is part of llvm not llvm-libs, so without it in llvm-svn we just don't have it.
As for the others, are we suppose to install the non-svn version to get them?

Thank you!

Last edited by gee (2016-03-09 08:48:22)

lordheavy · 2016-03-14 17:03:17

gee wrote:

@lordheavy: I'm running into an issue with llvm-svn from your repos, and I think it's related to this bug:
https://bugs.archlinux.org/task/46468
My issue is the same as the bug, I am trying to build a package with cmake, and it complains about the absence of LLVMHello.so.
also this line seems not good either:
rm -f "$pkgdir"/usr/lib/{libLLVM,libLTO,LLVMgold,BugpointPasses}.so
BugpointPasses is part of llvm not llvm-libs, so without it in llvm-svn we just don't have it.
As for the others, are we suppose to install the non-svn version to get them?
Thank you!

You need llvm-libs to get LTO, i've fixed LLVMHello and BugpointPasses.

++

gee · 2016-03-15 00:50:47

That worked for me, thank you!
I'm curious though, why did you make llvm-libs-svn not conflicting with llvm-libs? Are there packages that you do not build that depend on llvm-libs and so you keep them working that way?

Last edited by gee (2016-03-15 01:32:36)

lordheavy · 2016-03-19 22:25:03

gee wrote:

That worked for me, thank you!
I'm curious though, why did you make llvm-libs-svn not conflicting with llvm-libs? Are there packages that you do not build that depend on llvm-libs and so you keep them working that way?

The main goal to split llvm-libs and llvm is to be able to install several llvm-libs version (we have also llvm35-libs)

gee · 2016-03-21 02:16:55

I see thanks!

el_Salmon · 2016-03-31 23:15:47

What about the current OpenCL status? I have a HD 6490M (Northern Islands), but stock mesa driver doesn't support Image processing (that say "darktable -d opencl"). I give a try to mesa-git

Lone_Wolf · 2016-04-01 12:35:40

opencl-mesa relies on liblclc to provide the opencl functionality.

check /usr/lib/clc/ folder for a firmware that matches the chipset your card uses.

If there is none, you can try amd app sdk .
GPGPU page on wiki has more info.

Arch Linux

#2876 2015-12-23 13:55:18

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2877 2015-12-23 14:42:18

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2878 2015-12-23 15:00:08

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2879 2015-12-23 17:07:01

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2880 2015-12-24 12:37:27

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2881 2015-12-24 12:52:11

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2882 2015-12-24 14:17:06

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2883 2015-12-24 14:29:57

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2884 2015-12-24 15:30:48

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2885 2016-01-02 23:02:56

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2886 2016-01-03 16:52:27

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2887 2016-01-03 18:28:53

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2888 2016-01-25 19:55:41

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2889 2016-02-07 12:21:00

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2890 2016-02-07 20:13:45

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2891 2016-02-09 13:42:19

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2892 2016-02-12 10:37:38

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2893 2016-02-12 16:21:20

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2894 2016-03-09 08:30:22

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2895 2016-03-14 17:03:17

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2896 2016-03-15 00:50:47

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2897 2016-03-19 22:25:03

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2898 2016-03-21 02:16:55

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2899 2016-03-31 23:15:47

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2900 2016-04-01 12:35:40

Re: Discussion: the ati/radeon open source drivers & radeon repository

Board footer