Bon j'ai testé de nouveau Blender 2.80alpha et l'OpenCL, il ne manque plus que le rendu et c'est tout bon.
[USER... Blender-2-80]$ CYCLES_OPENCL_SPLIT_KERNEL_TEST=1 ./blender
Read prefs: /home/sylvain/.config/blender/2.80/config/userpref.blend
ATTENTION: default value of option vblank_mode overridden by environment.
Warning! Unable to find a multisample pixel format that supports exactly 16 samples. Substituting one that uses 8 samples.
found bundled python: /home/sylvain/3D/Experimental/Blender-2-80/2.80/python
Device init success
Device init success
Compiling OpenCL program split
Kernel compilation of split finished in 1.84s.
Compiling OpenCL program base
Kernel compilation of base finished in 1.21s.
Compiling OpenCL program split
OpenCL build failed with error CL_BUILD_PROGRAM_FAILURE, errors in console.
OpenCL program split build output: source/kernel/kernel_compat_opencl.h:146:9: warning: 'NULL' macro redefined
/usr/include/clc/clctypes.h:89:9: note: previous definition is here
source/util/util_math_float3.h:377:32: warning: double precision constant requires cl_khr_fp64, casting to single precision
source/util/util_math_float3.h:378:32: warning: double precision constant requires cl_khr_fp64, casting to single precision
source/util/util_math_float3.h:379:32: warning: double precision constant requires cl_khr_fp64, casting to single precision
source/util/util_math.h:517:14: warning: implicit declaration of function 'lgamma' is invalid in C99
source/kernel/closure/bsdf_ashikhmin_shirley.h:136:52: warning: implicit declaration of function 'native_tan' is invalid in C99
source/kernel/kernel_compat_opencl.h:128:19: note: expanded from macro 'tanf'
<unknown>:0:0: in function kernel_ocl_path_trace_data_init void (i8 addrspace(1)*, %struct.KernelData addrspace(2)*, i8 addrspace(1)*, i32, i8 addrspace(1)*, i32 addrspace(1)*, <4 x float> addrspace(1)*, <4 x float> addrspace(1)*, <4 x float> addrspace(1)*, i32 addrspace(1)*, i32 addrspace(1)*, i32 addrspace(1)*, i32 addrspace(1)*, i32 addrspace(1)*, i32 addrspace(1)*, <2 x float> addrspace(1)*, <4 x float> addrspace(1)*, <4 x float> addrspace(1)*, i32 addrspace(1)*, <4 x float> addrspace(1)*, <4 x i32> addrspace(1)*, i32 addrspace(1)*, <2 x float> addrspace(1)*, <4 x float> addrspace(1)*, <4 x float> addrspace(1)*, i32 addrspace(1)*, <4 x i32> addrspace(1)*, float addrspace(1)*, <4 x float> addrspace(1)*, <4 x i8> addrspace(1)*, <4 x float> addrspace(1)*, <4 x float> addrspace(1)*, <2 x float> addrspace(1)*, <2 x float> addrspace(1)*, <4 x float> addrspace(1)*, <4 x i32> addrspace(1)*, i32 addrspace(1)*, i32 addrspace(1)*, float addrspace(1)*, i32 addrspace(1)*, <4 x i8> addrspace(1)*, <4 x float> addrspace(1)*, i8 addrspace(1)*, float addrspace(1)*, <4 x i32> addrspace(1)*, i32, i32, i32, i32, i32, i32, i32, i32, i32 addrspace(1)*, i32, i8 addrspace(1)*, i32 addrspace(1)*, i32, float addrspace(1)*): unsupported call to function get_global_id
Device init success
Device init success
Compiling OpenCL program split
OpenCL build failed with error CL_BUILD_PROGRAM_FAILURE, errors in console.
OpenCL program split build output: source/kernel/kernel_compat_opencl.h:146:9: warning: 'NULL' macro redefined
/usr/include/clc/clctypes.h:89:9: note: previous definition is here
source/util/util_math_float3.h:377:32: warning: double precision constant requires cl_khr_fp64, casting to single precision
source/util/util_math_float3.h:378:32: warning: double precision constant requires cl_khr_fp64, casting to single precision
source/util/util_math_float3.h:379:32: warning: double precision constant requires cl_khr_fp64, casting to single precision
source/util/util_math.h:517:14: warning: implicit declaration of function 'lgamma' is invalid in C99
source/kernel/closure/bsdf_ashikhmin_shirley.h:136:52: warning: implicit declaration of function 'native_tan' is invalid in C99
source/kernel/kernel_compat_opencl.h:128:19: note: expanded from macro 'tanf'
<unknown>:0:0: in function kernel_ocl_path_trace_data_init void (i8 addrspace(1)*, %struct.KernelData addrspace(2)*, i8 addrspace(1)*, i32, i8 addrspace(1)*, i32 addrspace(1)*, <4 x float> addrspace(1)*, <4 x float> addrspace(1)*, <4 x float> addrspace(1)*, i32 addrspace(1)*, i32 addrspace(1)*, i32 addrspace(1)*, i32 addrspace(1)*, i32 addrspace(1)*, i32 addrspace(1)*, <2 x float> addrspace(1)*, <4 x float> addrspace(1)*, <4 x float> addrspace(1)*, i32 addrspace(1)*, <4 x float> addrspace(1)*, <4 x i32> addrspace(1)*, i32 addrspace(1)*, <2 x float> addrspace(1)*, <4 x float> addrspace(1)*, <4 x float> addrspace(1)*, i32 addrspace(1)*, <4 x i32> addrspace(1)*, float addrspace(1)*, <4 x float> addrspace(1)*, <4 x i8> addrspace(1)*, <4 x float> addrspace(1)*, <4 x float> addrspace(1)*, <2 x float> addrspace(1)*, <2 x float> addrspace(1)*, <4 x float> addrspace(1)*, <4 x i32> addrspace(1)*, i32 addrspace(1)*, i32 addrspace(1)*, float addrspace(1)*, i32 addrspace(1)*, <4 x i8> addrspace(1)*, <4 x float> addrspace(1)*, i8 addrspace(1)*, float addrspace(1)*, <4 x i32> addrspace(1)*, i32, i32, i32, i32, i32, i32, i32, i32, i32 addrspace(1)*, i32, i8 addrspace(1)*, i32 addrspace(1)*, i32, float addrspace(1)*): unsupported call to function get_global_id
Erreur: Failed loading render kernel, see console for errors
Saved session recovery to '/tmp/quit.blend'
Blender quit
Sur Fedora 26 + LLVM/CLANG 4.0.x + Mesa 17.2git