Package Details: catalyst-test 15.12-8

Git Clone URL: https://aur.archlinux.org/catalyst-test.git (read-only)
Package Base: catalyst-test
Description: AMD/ATI Catalyst drivers for linux AKA Crimson. catalyst-dkms + catalyst-utils + lib32-catalyst-utils + experimental powerXpress suppport. PRE-GCN Radeons ARE NOT SUPPORTED
Upstream URL: http://www.amd.com
Licenses: custom
Conflicts: catalyst, catalyst-daemon, catalyst-dkms, catalyst-generator, catalyst-hook, catalyst-utils, libcl, libgl, mesa-libgl, mesa-libgl-git
Provides: catalyst=15.12, catalyst-hook=15.12, catalyst-libgl=15.12, catalyst-utils=15.12, dri, libatical=15.12, libcl, libgl, libtxc_dxtn, mesa-libgl, mesa-libgl-git, opencl-catalyst=15.12
Submitter: Vi0L0
Maintainer: Vi0L0
Last Packager: Vi0L0
Votes: 168
Popularity: 0.482219
First Submitted: 2010-02-17 20:49
Last Updated: 2016-07-25 20:37

Dependencies (22)

Required by (679)

Sources (26)

  • 4.3-gentoo-mtrr.patch
  • 4.3-kolasa-seq_printf.patch
  • 4.4-manjaro-xstate.patch
  • 4.6-arch-get_user_pages-page_cache_release.patch
  • 4.7-arch-cpu_has_pge.patch
  • arch-fglrx-authatieventsd_new.patch
  • ati-powermode.sh
  • atieventsd.service
  • atieventsd.sh
  • catalyst.conf
  • catalyst.sh
  • crimson_i686_xg.patch
  • dkms.conf
  • fglrx_gpl_symbol.patch
  • grsec_arch.patch
  • http://www2.ati.com/drivers/linux/radeon-crimson-15.12-15.302-151217a-297685e.zip
  • lano1106_fglrx_intel_iommu.patch
  • lano1106_kcl_agp_13_4.patch
  • lib32-catalyst.sh
  • makefile_compat.patch
  • makesh-dont-check-gcc-version.patch
  • pxp_switch_catalyst
  • switchlibGL
  • switchlibglx
  • temp-links-catalyst.service
  • temp_links_catalyst

Latest Comments

synergy commented on 2016-07-30 01:55

Thank you.
When I rebooted, one of the vsync options was enabled again and I had to disable and reboot again. I now get performance closer to what I've come to expect in source games.

Re: compositing, I mainly use i3wm without a compositor, I just wanted to do additional testing.

Vi0L0 commented on 2016-07-29 07:08

In amdcccle I'm always disabling v-sync as well, it's not needed when using tear free.
There's also simple tear free switcher which allows switching via console (it comes with desktop entry): https://aur.archlinux.org/packages/amdtfset-git/

Also keep in mind that working composite has negative impact on performance of fullscreen rendering apps. I'm always disabling this as well before running app/game - in KDE its easy with Alt+Shift+F12 shortcut

synergy commented on 2016-07-29 01:22

tahiti, arch linux testing:
I installed, tested, and everything ran great except video playback (to be expected coming from radeon).

I then ran a full system upgrade (mesa updated), enabled tear free, disabled tear free (from cli), rebooted, and I now have extreme mouse lag and stutter in source games. Using the xorg 1.17 repo, fglrx is still loaded, not radeon (and opencl works).

Any ideas? I should be getting a few hundred fps in these games.

Edit: forgot to mention, I get "IDirect3DQuery9::GetData(): Occlusion query flush took xx.xxms!" spammed in console.

I've also tried toggling tear free again with no change, and tried in both i3wm and plasma with no change.

Edit Again: FIXED! I had to go to the "more settings" page in amdcccle as well to disable it separately there (weird layout). Performance is still way under the opensource driver (can run 8x msaa in open source driver, can't run any on fglrx for playable fps), but it's a trade off for opencl.

Vi0L0 commented on 2016-07-25 20:45

@irton: ah yes, will fix it later silently, thanks

for now I added 4.7 kernel support, it seems to work fine with my amateur patch.

errors that were generated without 4.7 patch were like (could be useful for searchers):
-----8<-------
/var/lib/dkms/fglrx/15.12/build/2.6.x/firegl_public.c: In function ?kcl_mem_pat_setup?:
./arch/x86/include/asm/bitops.h:337:29: error: passing argument 2 of ?constant_test_bit? from incompatible pointer type [-Werror=incompatible-pointer-types]
? constant_test_bit((nr), (addr)) \
^
/var/lib/dkms/fglrx/15.12/build/2.6.x/firegl_public.h:654:21: note: in expansion of macro ?test_bit?
#define cpu_has_pge test_bit(X86_FEATURE_PGE, &boot_cpu_data.x86_capability)
^~~~~~~~
/var/lib/dkms/fglrx/15.12/build/2.6.x/firegl_public.c:4547:9: note: in expansion of macro ?cpu_has_pge?
if (cpu_has_pge)
^~~~~~~~~~~
./arch/x86/include/asm/bitops.h:308:28: note: expected ?const volatile long unsigned int *? but argument is of type ?__u32 (*)[19] {aka unsigned int (*)[19]}?
static __always_inline int constant_test_bit(long nr, const volatile unsigned long *addr)
^~~~~~~~~~~~~~~~~
./arch/x86/include/asm/bitops.h:338:29: error: passing argument 2 of ?variable_test_bit? from incompatible pointer type [-Werror=incompatible-pointer-types]
: variable_test_bit((nr), (addr)))
^
/var/lib/dkms/fglrx/15.12/build/2.6.x/firegl_public.h:654:21: note: in expansion of macro ?test_bit?
#define cpu_has_pge test_bit(X86_FEATURE_PGE, &boot_cpu_data.x86_capability)
^~~~~~~~
(...)
-----8<-------

irton commented on 2016-07-04 05:29

script make.sh may be with execute permission? eq. 755

Vi0L0 commented on 2016-05-30 18:42

15.12-7: fixed dkms versioning
Also noticed:
- no more wh problems on cs:go on 4.5 and 4.6 kernels - probably valve fixed it;
- for some reason glxgears32 segfaults on my pc (no matter of what catalyst version, didnt check oss amdgpu/radeon):
'glxgears32' terminated by signal SIGSEGV (Address boundary error)
will check it later;
- tomorrow I should be able to update repos, at least [catalyst]

Vi0L0 commented on 2016-05-13 20:55

thanks again :) so it stores modules in two different places... oukaaaay 8)
Arch/dkms wiki should definitelly have more details

runnytu commented on 2016-05-13 20:26

@Vi0L0, no problem we are all humans after all ;P, the best way to clean is 'rm -fr /var/lib/dkms/fglrx' and to install the module for all the installed kernels with 'dkms autoinstall -m fglrx-15.12 -v 6' later with 'dkms status' should see a output like this for all installed kernels, 2 in my case:
fglrx-15.12, 6, 4.5.4-1-ARCH, x86_64: installed
fglrx-15.12, 6, 4.5.4-1-zen, x86_64: installed

Vi0L0 commented on 2016-05-13 19:48

you got the point runnytu, nice finding

I have removed dkms commands from install script, should be fine now.
Previous version was taken from original catalyst-dkms version and it didn't fit arch's newer dkms.
It was creating fglrx.ko module for booted kernel (with "dkms -v fglrx"), then it was overwritten (because defined module name is the same) by dkms itself (with "dkms -m fglrx-15.12").

I'm just not sure how is dkms going to find out about overwritten module (as it was generated using different version, could generate some trash from it that will have to be removed manually. Sorry for the inconvinience.
Possibly the best way to clean after -5 version is to:
1. use `dkms remove -m fglrx-15.12 -v 5 --all` or `dkms remove -m fglrx-15.12 -v 6 --all` after update to -6
2. use `find /usr/lib/modules/ -iname '*fglrx*'` to trace trash and use rm to remove it manually
3. use `dkms install -m fglrx-15.12 -v 6` to build module for booted kernel or reinstall catalyst-test if you got more kernels - I don't know simple way to build modules for all kernels. One can use some loop like `for` for it, but there should be some easier way available.
But even without cleaning it should work fine, eventually there will be some fglrx module trash left after kernel's version bump ;P.

Thanks again runnytu.
And sorry for missing it.

runnytu commented on 2016-05-12 18:04

Hi, @Vi0L0, you should remove the 'dkms install -m fglrx -v $1' in post_install() and post_upgrade() and the 'dkms remove -m fglrx -v $1 --all' in pre_remove() in the file catalyst-test.install because pacman installs and removes the DKMS modules automatically at the beginning, to remove and at the end to install, with this commands installs the DKMS modules twice, one with your scipt, named fglrx that fails sometimes and the automatic pacman, named fglrx-15.12 that works fine, could you check it?

All comments