AMD OpenCL support and Blender segfault

[KAMIKAZE_]

Hello. I have AMD Radeon HD 7970 videocard. Finally got fglrx working (btw KDE and it's apps just freeze, any other like xfce/gnome - just work fine).
Is there anyone who succeeded with OpenCL setup and Blender Cycles rendering. Right now I can't open settings window or even switch rendering engine to cycles - just getting:

• [ 130.815838] blender-bin-2.6[3163]: segfault at 1000000000c ip 00000037280a1e26 sp 00007fffe119ddb0 error 6 in libstdc++.so.6.0.14[3728000000+f1000]

clinfo shows the same:

• [ 1131.431450] clinfo[3381]: segfault at 1000000000c ip 00000037280a1e26 sp 00007fffebce5d90 error 6 in libstdc++.so.6.0.14[3728000000+f1000]

[earthy]

Yeah I did have cycles working on 4 boxes.But after upgrade they stopped working on 2.I have been trying various versions on them to get them back up.I have converted most of my projects to cycles,and they always render much nicer.I am rendering on those 2 working box,turned one off(no work for it),and am trying to fix blender on 1.Last one I tried (2.63a-r2::aluco) did not seem to install correctly,as I can not find the command(supposed to be something like blender-2.62).Blender-2.63a-r1 installed correctly and renders a small test scene,but none of my projects.I was looking for a aluco forum of some sort when I found this post.The boxes with the working cycles have:2.62-r2 and 2.63a ,both from aluco overlay.and btw only the 1 with 2.62-r2 has a screen attached to it, so the others I run from console.

[earthy]

Using these flags:cycles dds ffmpeg game-engine nls player sdl sse
and now it finishes emerge on -2.62-r2 but gives me these :

Code: Select all

* Remaining bundled dependencies:
 * - extern/lzo
 * - extern/lzma
 * - extern/binreloc
 * - extern/carve
 * - extern/libmv
 * - extern/recastnavigation
 * - extern/eltopo
 * - extern/libredcode

and render fails :

Code: Select all

Fra:1 Mem:25.71M (84.47M, peak 217.78M) | Elapsed: 40:07.24 | Updating Scene BVH | Building BVH 100%, duplicates 0%
Fra:1 Mem:25.71M (84.47M, peak 217.78M) | Elapsed: 40:13.93 | Updating Scene BVH | Packing BVH triangles
terminate called after throwing an instance of 'std::bad_alloc'
  what():  std::bad_alloc
Aborted

Testing of these takes quite some time as I have given up on the test scene(doesn´t give satifactory estimate of operability of cycles),and as u see this render fails at 40 minutes,just before starting the actual path-tracing(rendering).The other boxes really want to upgrade to 2.63a,but I won´t let them .
1 of the working boxes has these flags

Code: Select all

 2.63a(2.6)[2](01:40:41 07/23/12)(cycles dds elbeem ffmpeg fftw game-engine openexr openmp player sdl sse

[earthy]

Got another of them working from aluco overlay:

Code: Select all

 2.62(2.60)[2](00:36:54 07/26/12)(cycles dds elbeem ffmpeg game-engine openexr openmp sdl sse

the fourth one succeeded on cyclestest.blend and is now rendering a big one.It has the same version as the 1 above.Those 2,that were not working,are x86 the other ones amd64.

[earthy]

Both of them failed at further tests,one withe a segfault
I installed emerge =media-gfx/blender-2.63a-r2::gentoo but can't find command.just blender,which looks like blender2.49b.

[earthy]

The command for media-gfx/blender-2.63a-r2::gentoo seems to be just blender.However it looks like it can't render a .blend made for cycles despite having a cycles flag.I am trying this on one of the non-working boxes in CLI.Here is a part of the output of that failed render(renders just a black picture):

Code: Select all

RNA_string_set: OperatorProperties.data_path not found.
RNA_boolean_set: OperatorProperties.reverse not found.
search for unknown operator 'WM_OT_context_cycle_int', 'WM_OT_context_cycle_int'
RNA_string_set: OperatorProperties.data_path not found.
RNA_boolean_set: OperatorProperties.reverse not found.
search for unknown operator 'WM_OT_context_cycle_int', 'WM_OT_context_cycle_int'
RNA_string_set: OperatorProperties.data_path not found.
RNA_boolean_set: OperatorProperties.reverse not found.
search for unknown operator 'CONSOLE_OT_execute', 'CONSOLE_OT_execute'
search for unknown operator 'CONSOLE_OT_execute', 'CONSOLE_OT_execute'
search for unknown operator 'CONSOLE_OT_autocomplete', 'CONSOLE_OT_autocomplete'
search for unknown operator 'WM_OT_context_toggle_enum', 'WM_OT_context_toggle_enum'
RNA_string_set: OperatorProperties.data_path not found.
RNA_string_set: OperatorProperties.value_1 not found.
RNA_string_set: OperatorProperties.value_2 not found.
search for unknown operator 'WM_OT_context_toggle', 'WM_OT_context_toggle'
RNA_string_set: OperatorProperties.data_path not found.
search for unknown operator 'WM_OT_context_toggle', 'WM_OT_context_toggle'
RNA_string_set: OperatorProperties.data_path not found.
search for unknown operator 'WM_OT_context_toggle', 'WM_OT_context_toggle'
RNA_string_set: OperatorProperties.data_path not found.
search for unknown operator 'WM_OT_context_toggle', 'WM_OT_context_toggle'
RNA_string_set: OperatorProperties.data_path not found.
search for unknown operator 'WM_OT_context_toggle', 'WM_OT_context_toggle'
RNA_string_set: OperatorProperties.data_path not found.
read blend: /home/earthy/cyclestest.blend
Fra:1 Mem:7.31M (0.10M, peak 8.71M) | Preparing Scene data
Fra:1 Mem:26.18M (0.10M, peak 80.77M) | Preparing Scene data
Fra:1 Mem:26.63M (0.10M, peak 80.77M) | Preparing Scene data
Fra:1 Mem:26.63M (0.10M, peak 80.77M) | Creating Shadowbuffers
Fra:1 Mem:26.63M (0.10M, peak 80.77M) | Raytree.. preparing
Fra:1 Mem:54.67M (0.10M, peak 80.77M) | Raytree.. building
Fra:1 Mem:55.22M (0.10M, peak 100.29M) | Raytree finished
Fra:1 Mem:55.22M (0.10M, peak 100.29M) | Creating Environment maps
Fra:1 Mem:55.22M (0.10M, peak 100.29M) | Caching Point Densities
Fra:1 Mem:55.22M (0.10M, peak 100.29M) Sce: Scene Ve:244307 Fa:245021 La:0
Fra:1 Mem:55.22M (0.10M, peak 100.29M) | Loading voxel datasets
Fra:1 Mem:55.22M (0.10M, peak 100.29M) Sce: Scene Ve:244307 Fa:245021 La:0
Fra:1 Mem:55.22M (0.10M, peak 100.29M) | Occlusion preprocessing
Fra:1 Mem:71.75M (0.10M, peak 100.29M) | Volume preprocessing
Fra:1 Mem:71.75M (0.10M, peak 100.29M) Sce: Scene Ve:244307 Fa:245021 La:0
Fra:1 Mem:71.75M (0.10M, peak 100.29M) Sce: Scene Ve:244307 Fa:245021 La:0
write exr tmp file, 1960x1600, /tmp/cyclestest.blend_Scene.exr
Fra:1 Mem:94.71M (21.41M, peak 138.99M) | Scene, Part 2-16
Fra:1 Mem:94.71M (21.41M, peak 139.09M) | Scene, Part 1-16
Fra:1 Mem:71.75M (21.41M, peak 139.09M) | Scene, Part 3-16
Fra:1 Mem:94.62M (21.41M, peak 139.09M) | Scene, Part 4-16
Fra:1 Mem:94.62M (21.41M, peak 139.09M) | Scene, Part 6-16
Fra:1 Mem:94.71M (21.41M, peak 139.09M) | Scene, Part 5-16
Fra:1 Mem:94.62M (21.41M, peak 139.09M) | Scene, Part 7-16
Fra:1 Mem:94.71M (21.41M, peak 139.09M) | Scene, Part 8-16
Fra:1 Mem:94.71M (21.41M, peak 139.09M) | Scene, Part 9-16
Fra:1 Mem:94.62M (21.41M, peak 139.09M) | Scene, Part 10-16
Fra:1 Mem:94.71M (21.41M, peak 139.09M) | Scene, Part 11-16
Fra:1 Mem:71.75M (21.41M, peak 139.09M) | Scene, Part 13-16
Fra:1 Mem:81.76M (21.41M, peak 139.09M) | Scene, Part 12-16
Fra:1 Mem:86.26M (21.41M, peak 139.09M) | Scene, Part 15-16
Fra:1 Mem:94.62M (21.41M, peak 139.09M) | Scene, Part 16-16
Fra:1 Mem:71.75M (10.76M, peak 139.09M) | Scene, Part 14-16
read exr tmp file: /tmp/cyclestest.blend_Scene.exr
Fra:1 Mem:7.19M (167.58M, peak 239.33M) Sce: Scene Ve:244307 Fa:245021 La:0
Saved: /home/earthy/renders/0001.png Time: 13:28.71

Blender quit

This looks like a non-cycles render.
For comparision a working cycles render output:

Code: Select all

Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.19 | Loading render kernels (may take a few minutes the first time)
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.19 | Waiting for render to start
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.22 | Updating Scene
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.22 | Updating Background
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.22 | Updating Shaders
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.22 | Updating Images
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.22 | Updating Camera
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.22 | Updating Objects
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.22 | Updating Objects | Copying Transformations to device
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.25 | Updating Objects | Applying Static Transformations
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.25 | Updating Meshes
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.27 | Updating Mesh | Computing normals
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.27 | Updating Mesh | Copying Mesh to device
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.27 | Updating Mesh | Computing attributes
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.27 | Updating Mesh | Copying Attributes to device
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.27 | Updating Mesh BVH Cube.577 68/3 | Building BVH
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.28 | Updating Mesh BVH Cube.577 68/3 | Packing BVH triangles
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.28 | Updating Mesh BVH Cube.577 68/3 | Packing BVH nodes
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.28 | Updating Mesh BVH Cube.297 347/3 | Building BVH
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.28 | Updating Mesh BVH Cube.297 347/3 | Packing BVH triangles
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.28 | Updating Mesh BVH Cube.297 347/3 | Packing BVH nodes
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.28 | Updating Mesh BVH Cube 631/3 | Building BVH
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.28 | Updating Mesh BVH Cube 631/3 | Packing BVH triangles
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.28 | Updating Mesh BVH Cube 631/3 | Packing BVH nodes
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.28 | Updating Scene BVH | Building
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.28 | Updating Scene BVH | Building BVH
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.58 | Updating Scene BVH | Building BVH 0%, duplicates 0%
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:00.84 | Updating Scene BVH | Building BVH 1%, duplicates 0%
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:01.34 | Updating Scene BVH | Building BVH 3%, duplicates 0%
Fra:442 Mem:357.22M (18.41M, peak 698.51M) | Elapsed: 00:01.65 | Updating Scene BVH

So I was wondering if blender is the command for 2.63a-r2::gentoo,what would be the command for blender-2.49b,which is the last version of old blender,which I know a lot of people like to have around.I might try installing that on 1 of the failing boxes.

[earthy]

I made an even simpler testfile.no cycles,no AO,no indirect lighting.1 light,1 camera 1 object.The output from CLI looks normal but the render is black.I am updating that box now.I bet I will not try 2.63a-r2::gentoo again.It is totally useless except for heating up the processor.

[jathlon]

There are some suggestions for using latest versions of blender with AMD cards in this forum thread over at Blender Artists;

http://blenderartists.org/forum/showthr ... rds-owners

The short answer is that opencl and amd cards are not working very well/at all with blender and cycles.

The amd 79xx series cards, spec wise, appear to potentially have much greater gpu compute possibilities than nvidia. Saddly it is not realized in blender.

[earthy]

Yeah I know I can't use ati cards for gfx card renders although I thought U could use them to do cpu renders.Anyways, my gfx cards are all too old to support CUDA.I have to carefully balance the frequency of updates, with my need to have boxes rendering,nonetheless I have 1 box remaining rendering and I need to update that.........but I found this thread,hoping that would fix box 2.
http://translate.googleusercontent.com/ ... WwyWhVjXaQ
and it doesn´t.

[earthy]

I have been using fedora in virtualbox and win7, just to get along but finally had 2.63a working from overlay aluco on box 2.
but not on box 1.
Am writing this on winblows, but I remember it was some error concerning libboost.
will post l8r with full details.