-
-
Notifications
You must be signed in to change notification settings - Fork 4.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Build error PCL + VTK 6.1 + CUDA #776
Comments
Lets continue my issues #776 here My system :
We have VTK 6.1 in common so I think its only this migration problem you mention. But you are mentioning the CUDA section build and not the gpu section build. I tought the CUDA section was depreciated a long time ago and only the gpu section was needed. Kinfu is in the gpu section. Am I wrong somewhere ? Anyway , cuda or gpu section , the problem come from VTK. In pcl_gpu_features_generated_centroid.cu.obj.cmake on line 79 we can find this related to VTK -DvtkIOGeometry_AUTOINIT=1(vtkIOMPIParallel); and I can find similars line on every cmake on the gpu section (features, containers, kinfu .) As duga3 propose , I think we need to reformat the way VTK put theses definitions. But its beyond my programming capabilities for now |
@HaArLiNsH Seems that your issue is also caused by the VTK definitions. I just had a look at the PCL CMake scripts. |
Hi @duga3 your solution sounds plausible. Can you send a pull request with it? Thanks! |
Hi duga3, |
I have unfortunately no fast fix, but I think you have to find out what variables are passed to the nvcc compiler. I wrote that fix 3 years ago, I guess that cmake/nvcc handling changed a bit since then. |
I get the same errors, tried with VERBOSE = 1. How to change MARCO with a function? Kindly let me know. cuda/features/CMakeFiles/pcl_cuda_features.dir/build.make:977: recipe for target 'cuda/features/CMakeFiles/pcl_cuda_features.dir/src/pcl_cuda_features_generated_normal_3d.cu.o' failed |
this caused by "add_definitions(${PCL_DEFINITIONS})"
this works for me |
I'm trying to build the latest master with VTK 6.1 and GPU support.
This might relate to #766.
I used the following to configure:
My system:
All libs are correctly found and the configuration finishes without errors.
Compiling stops with:
The error is caused from line 79 in "pcl_cuda_features_generated_normal_3d.cu.o.cmake".
The VTK cmake files add definitions like
They end up in the
nvcc_flags
variable and cause nvcc to crash.When I removed them manually, the cuda file I've done it for, compiles.
These definitions come from the new way VTK6 initializes the object factories.
nvcc seems like not to like these and exits with
The VTK cmake files automatically add these vtk* definitions.
More details here: http://www.vtk.org/Wiki/VTK/VTK_6_Migration/Factories_now_require_defines
For another project I solved this problem by adding a "vtk definitions filter" to the module that compiles the CUDA files.
Might no be the best solution, but the cmake code looks like this.
The text was updated successfully, but these errors were encountered: