ATI R600/700 OSS 3D Driver Reaches Gears Milestone

[FarJumper]

I still can't get it running on my Karmic with kernel 2.6.30. Mainline kernels don't work (glxgears returns -22 error). So I decided to build latest drm modules by easy-drm-modules-installer script. Still no luck. I got a lot of messages like:

[ 32.045787] radeon: disagrees about version of symbol drm_open
[ 32.045790] radeon: Unknown symbol drm_open
[ 32.045861] radeon: disagrees about version of symbol drm_fasync
[ 32.045863] radeon: Unknown symbol drm_fasync
[ 32.045933] radeon: disagrees about version of symbol drm_poll
[ 32.045935] radeon: Unknown symbol drm_poll
[ 32.046004] radeon: disagrees about version of symbol drm_get_resource_len
[ 32.046007] radeon: Unknown symbol drm_get_resource_len
[ 32.046096] radeon: disagrees about version of symbol drm_core_get_reg_ofs
[ 32.046098] radeon: Unknown symbol drm_core_get_reg_ofs
[ 32.046207] radeon: disagrees about version of symbol drm_irq_uninstall
[ 32.046210] radeon: Unknown symbol drm_irq_uninstall
[ 32.046286] radeon: disagrees about version of symbol drm_get_dev

[forum1793]

With changes made in last ~12 hours, hd3200 now gets 900+ fps in glxgears. eduke32 still goofed up. The changes, most likely in agd5f's drm 3D seemed to have helped hd3200 performance.

[tettamanti]

I still can't get it running on my Karmic with kernel 2.6.30. Mainline kernels don't work (glxgears returns -22 error). So I decided to build latest drm modules by easy-drm-modules-installer script. Still no luck. I got a lot of messages like:

[ 32.045787] radeon: disagrees about version of symbol drm_open
[ 32.045790] radeon: Unknown symbol drm_open

I'm not familiar with the script that you mention, but these messages mean that the headers used to compile the modules are from a kernel different from the one running.

[Kano]

I would use my script

http://kanotix.com/files/fix/dkms/drm-r6xx-r7xx-3d-dkms.sh

[suokko]

Is it normal to have direct rendering YES even though I don't have 3D capable driver yet ?
If I read through glxinfo's output I get a GL_RENDERER software rasterizer, how can I be sure to have 3D acceleration if glxinfo somehow fails to report correct things ?
Is glxinfo still relevant ?

Software rendering is direct too. Old guides saying to look if glxinfo says direct or no direct rendering are all outdated. Now you have to do glxinfo |grep OpenGL to see if renderer is Mesa DRI. (It is also good idea to check that correct mesa version is printed if you have problems.)

[nanonyme]

Software rendering is direct too. Old guides saying to look if glxinfo says direct or no direct rendering are all outdated.

and were wrong to begin with

[lucky_]

Ok thx, I was really suspicious about those guides, and GL RENDERER "software raserization" sounded really "not hardware acceleration for you".

[nanonyme]

Ok thx, I was really suspicious about those guides, and GL RENDERER "software raserization" sounded really "not hardware acceleration for you".

Yeah, software rasterizer does considerably advanced software rendering. Currently it's the only way with an ATi card to get OpenGL 2.1 with open drivers. (albeit running very very slowly)

[pingufunkybeat]

Am I correct in understanding that you can use the kernel part of the drm (the one from the latest 2.6.31) , and that all the work is now going into libdrm + mesa + radeon?

Or is the kernel drm component (the one that supports 600/700 3D calls) that I believe was merged in recently also in a state of flux?

[nanonyme]

Am I correct in understanding that you can use the kernel part of the drm (the one from the latest 2.6.31) , and that all the work is now going into libdrm + mesa + radeon?

Or is the kernel drm component (the one that supports 600/700 3D calls) that I believe was merged in recently also in a state of flux?

The kernel DRM components haven't been merged into anything, you still have to use agd5f's repo for those. It is considered possible the stuff would be entering 2.6.32.