Package Details: amdgpu-pro-oglp 24.10_1787253-1

Git Clone URL: https://aur.archlinux.org/amdgpu-pro-installer.git (read-only, click to copy)
Package Base: amdgpu-pro-installer
Description: AMDGPU Pro OpenGL driver
Upstream URL: https://www.amd.com/en/support/kb/release-notes/rn-amdgpu-unified-linux-22-40
Licenses: custom: AMDGPU-PRO EULA
Groups: Radeon_Software_for_Linux
Provides: libgl
Submitter: LEW21
Maintainer: Ashark (mesmer)
Last Packager: mesmer
Votes: 125
Popularity: 1.09
First Submitted: 2016-03-28 22:44 (UTC)
Last Updated: 2024-07-02 19:24 (UTC)

Dependencies (2)

Required by (744)

Sources (18)

Pinned Comments

mesmer commented on 2022-12-23 04:31 (UTC) (edited on 2023-05-28 19:28 (UTC) by mesmer)

IF PROBLEM HAPPEN

this package is our try to put amdgpu in ARCH, but this could break or have problems as amd update and change their packages for ubuntu

if anything happens you can always download a older tag in https://github.com/Ashark/archlinux-amdgpu-pro/releases and get the PKGBUILD there, and use makepkg with that build while we update or try to fix :)

ALWAYS USE LINUX-FIRMWARE-GIT IF LATEST DRIVER FAIL BEFORE ISSUE SOME ERROR

https://aur.archlinux.org/packages/linux-firmware-git

Ashark commented on 2019-12-09 20:52 (UTC) (edited on 2022-12-06 08:45 (UTC) by Ashark)

PKGBUILD generator is located here: https://github.com/Ashark/archlinux-amdgpu-pro

AMDGPU PRO wiki page: https://wiki.archlinux.org/index.php/AMDGPU_PRO

New versions will appear here: https://repo.radeon.com/amdgpu/.

Latest Comments

« First ‹ Previous 1 .. 53 54 55 56 57 58 59 Next › Last »

Themaister commented on 2016-05-19 08:42 (UTC) (edited on 2016-05-19 09:46 (UTC) by Themaister)

Hm, still doesn't work. Here's a paste of Xorg.0.log. I'm using a GCN 1.1 card with CIK enabled in kernel if that helps. ldconfig is pointing to the right place. It seems like it's failing something, trying fall back to software which doesn't work. http://pastebin.com/bBNvgWAe Plain MESA with AMDGPU is working however.

Corngood commented on 2016-05-18 15:02 (UTC)

Sounds like the problem I had before I had ld.so.conf set up properly. Maybe 'ldconfig -p | grep libGL.so' and make sure it's finding the amdgpu-pro one rather than mesa? You could try the PKGBUILD from my pull request as it fixes the ld.so.conf. https://github.com/LEW21/archlinux-amdgpu/pull/1

Themaister commented on 2016-05-18 14:57 (UTC)

I'm also struggling to get GLX working at all. Either I get a segfault if I enable DRI3 or I don't get GLX at all ... Do you need to do something special to get GLX working?

Corngood commented on 2016-05-18 14:16 (UTC)

@libercv Yeah, I noticed that when I went through and fixed up the PKGBUILD. Now it makes a link for the ld and modprobe configs. PR is up at https://github.com/LEW21/archlinux-amdgpu/pull/1 if anyone wants to take a look. @Themaister Yeah, I have that problem too. It was broken by this change in the loader: https://github.com/KhronosGroup/Vulkan-LoaderAndValidationLayers/commit/c666cff44f25c1db0451f52818807683b0346f00 which was merged in 1.0.8. If I use 1.0.6 I get the extension, and it seems to work fine. It's a bug in the AMD vulkan ICD, which only reports VK_KHR_swapchain and VK_KHR_xbc_surface. I guess it worked by accident before that loader change. I tried reporting it on community.amd.com, but I'm still waiting for my post to be approved. If anyone has a better way of reporting bugs to them, let me know.

Themaister commented on 2016-05-18 14:07 (UTC)

For some reason, VK_KHR_surface is missing for me (VK_KHR_xcb_surface is present though). Anyone else having same issue?

libercv commented on 2016-05-18 10:38 (UTC)

Thanks a lot @Corngood, it works!!! Just one minor change. I copied /usr/lib/amdgpu-pro/ld.conf to /etc/ld.so.conf.d/01-amdgpu-pro.conf instead of creating a link to gbm. I couldn't make it work with the symbolic link, but the file already includes the path to gbm.

Corngood commented on 2016-05-17 21:24 (UTC)

@jhackler I'll try to fix up the PKGBUILD to do all this stuff when I get a chance, but for now: - Before you do this, consider making a snapshot or backup, and I suggest 'systemctl set-default multi-user' to avoid having to touch your kernel args to stop Xorg from running in case something goes wrong.. - 380 is GCN 1.2 I think, so you _should_ get the amdgpu driver if you - Update your kernel (I used 4.5.4) - Blacklist the radeon driver (put "blacklist radeon" in /etc/modprobe.d/blacklist-radeon.conf) - Reboot and verify amdgpu is loaded (lsmod | grep amdgpu) - At this point if you have mesa installed you should have a working system with GL, etc. If you are in text mode, 'systemctl isolate graphical' to start X, and if you want, run glxgears or something to verify - 'systemctl isolate multi-user' to quit Xorg if you launched it - Make sure you don't have anything driver specific in /etc/X11/xorg.conf, /etc/X11/xorg.conf.d/*, or /usr/share/X11/xorg.conf.d/* - Install this AUR package (e.g. 'pacaur -S --force amdgpu-pro') - I'm not sure if --force works with pacaur, but the point is to force overwrite the conflicting firmware files. If not, install it with 'pacman -S --force'. - 'sudo ln -s amdgpu-pro/gbm/gbm_amdgpu.so /usr/lib/x86_64-linux-gnu/gbm_amdgpu.so' - Alternatively you _might_ be able to make the ink in the amdgpu-pro dir, or add gbm/ to ld.conf, but this is what I did. Without it, Xorg wouldn't start. - 'sudo ln -sf 1.18 /usr/lib/x86_64-linux-gnu/amdgpu-pro/xorg' - The package ships with a symlink to the Xorg 1.15 driver, so you need this if you have an up to date 'xorg-server'. - create /etc/ld.so.conf.d/01-amdgpu-pro.conf with: ============== /usr/lib/x86_64-linux-gnu/amdgpu-pro /usr/lib/i386-linux-gnu/amdgpu-pro ============== - `sudo ldconfig` - Now you should be able to 'systemctl isolate graphical' and have everything working. - Steam does some shenanigans with the LD path, so I have to run it with "LD_LIBRARY_PATH=/usr/lib/i386-linux-gnu/amdgpu-pro/ steam". This could be fixed by including top level symlinks to libGL etc. If you have any problems feel free to email me. I wrote this from memory, so I could have forgot something. I'm going to go see how much of this I can fix in the PKGBUILD, which should make the process less painful.

Jhackler commented on 2016-05-17 09:11 (UTC)

Corngood If you don't mind, and have the time, could you post slightly more detailed instructions on how you did this, just so I can test it on my r9 380? If not I will eventually go through and figure it out myself or wait for a package update :/.

Corngood commented on 2016-05-15 19:39 (UTC)

It works! I just didn't have the amdgpu-pro libs in ld.so.conf. After adding the path, Xorg starts, GL works, Vulkan works. I haven't tested lib32 GL, VDPAU etc, but it's good enough for me to keep working on my vulkan projects after moving from NV to AMD. Also I tried it both on linux 4.2 + amdgpu-pro-dkms, and 4.5.4 + CIK. 4.2 doesn't detect my screens properly, but 4.5 does. I haven't noticed any other differences. So I guess the only real problems with this package on latest Arch are: - xorg 1.18 symlink - gbm_amdgpu.so symlink ? - ld.so.conf entry - firmware conflicts Then for people with GCN < 1.2 chips, getting the correct kernel (4.5.4 with CIK option or whatever GCN 1.0 patches AMD released in the last couple of days).

Corngood commented on 2016-05-14 16:32 (UTC)

Has anyone actually got this to work? I'm just going through the process now on my R9 290. I made some notes: - install amdgpu-pro - firmware conflict (mentioned below). I just overwrote with the amdgpu images from this package - built kernel 4.5.3 with CIK - downgraded to xorg 1.15 - crash in Xorg at GLX init - upgrade to xorg 1.18 - symlink xorg driver 1.18 (mentioned below) - similar crash in Xorg - build kernel 4.2 without amdgpu - install amdgpu-pro-dkms - OS_VERSION problem with /etc/os-release, fixed by defaulting OS_VERSION to 0.0 in the Makefile - -Wno- some warnings to make it compile with -Werror - /boot/System.map-[uname] is needed, so copied that from my kernel build - dkms module seems to load ok, but I still get a similar crash in Xorg - build debug Xorg - Xorg log now shows the signal handler as the crash stack - running in gdb shows this stack for the segfault: http://pastebin.com/HBVDpimZ So it's a crash during dlopen() of amdgpu_dri.so (calling 0), for which I have no symbols. I couldn't find anything relevant in strace or the xorg log before the crash. P.S. Thanks for packaging this. I probably wouldn't have bothered to try if it didn't exist.

« First ‹ Previous 1 .. 53 54 55 56 57 58 59 Next › Last »