Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

Problem with several OpenCL implementations and blender

hey,

I am trying to get blender to use my nVidia GPU, instead of the Intel one. I managed to start glxgears with the nVidia GPU (so, I am one step closer than last time I tried). However, when I start blender and open the system preferences, I get

: CommandLine Error: Option 'enable-value-profiling' registered more than once! LLVM ERROR: inconsistency in registered CommandLine options

and blender quits.

This is a known bug. The pocl authors discuss it here: https://github.com/pocl/pocl/issues/474 LLVM developers discuss it here: https://bugs.llvm.org/show_bug.cgi?id=30587

If I understand the problem correctly, several OpenCL implementations are being loaded in order to check which one is the best to use. All of them load LLVM, which leads to problems because the command line option 'enable-value-profiling' is registered once for each OpenCL implementation. Thereby this will happen for every program which uses OpenCL.

I have icd files for intel-beignet, mesa and pocl.

None of the bug reports I linked mentioned a work around, however, given that both threads are about 5 month old, I can not imagine that no one fixed the problem. So, is there a workaround? Can I uninstall some OpenCL implementations? If so, which?