nvtx definitions redefined

[LourensVeen]

When compiling Bonsai against CUDA 12.6, I'm getting

In file included from /var/scratch/lveen/ReDiTSAp/amuse/src/amuse/community/bonsai2/src/bonsai/runtime/src/sort_bodies_gpu.cu:2:
/var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/nvToolsExt.h:187: warning: "NVTX_DECLSPEC" redefined
  187 |     #define NVTX_DECLSPEC
      | 
In file included from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/cub/detail/nvtx3.hpp:635,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/cub/detail/nvtx.cuh:49,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/cub/device/device_for.cuh:40,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/system/cuda/detail/for_each.h:42,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/system/detail/adl/for_each.h:50,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/detail/for_each.inl:30,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/for_each.h:266,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/system/detail/generic/copy.inl:29,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/system/detail/generic/copy.h:51,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/detail/copy.inl:30,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/detail/copy.h:72,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/detail/allocator/copy_construct_range.inl:30,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/detail/allocator/copy_construct_range.h:45,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/detail/contiguous_storage.inl:29,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/detail/contiguous_storage.h:190,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/detail/temporary_array.h:47,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/system/cuda/detail/internal/copy_cross_system.h:47,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/system/cuda/detail/copy.h:79,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/system/cuda/detail/assign_value.h:33,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/system/detail/adl/assign_value.h:50,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/detail/reference.h:35,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/memory.h:35,
                 from /var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/thrust/device_ptr.h:33,
                 from /var/scratch/lveen/ReDiTSAp/amuse/src/amuse/community/bonsai2/src/bonsai/runtime/include/my_cuda_rt.h:27,
                 from /var/scratch/lveen/ReDiTSAp/amuse/src/amuse/community/bonsai2/src/bonsai/runtime/include/octree.h:21,
                 from /var/scratch/lveen/ReDiTSAp/amuse/src/amuse/community/bonsai2/src/bonsai/runtime/src/sort_bodies_gpu.cu:1:
/var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/nvtx3/nvToolsExt.h:207: note: this is the location of the previous definition
  207 | #define NVTX_DECLSPEC NVTX_INLINE_STATIC
      | 
/var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/nvToolsExt.h:223: warning: "NVTX_VERSION" redefined
  223 | #define NVTX_VERSION 2
      | 
/var/scratch/lveen/miniforge3/envs/amuse-dev/targets/x86_64-linux/include/nvtx3/nvToolsExt.h:184: note: this is the location of the previous definition
  184 | #define NVTX_VERSION 3
      | 

It seems like I have two copies of this nvToolsExt.h header installed, one in include/ and one in include/nvtx3/, and it seems that the former is included directly from sort_bodies_gpu.cu while the latter is included indirectly via thrust/device_ptr.h from my_cuda_rt.h.

So I tried replacing #include "nvToolsExt.h" in sort_bodies_gpu.cu with #include "nvtx3/nvToolsExt.h" and that seems to actually fix the problem, I can build the latest version with that change and the AMUSE stuff and it even passes the tests. I don't know enough about CUDA to know if that's a proper fix though.

[jbedorf]

This should not be an issue. That library is only used for tracking purposes.