- [CCCL headers moved to a different directory in CUDA 13](https://developer.nvidia.com/blog/whats-new-and-important-in-cuda-toolkit-13-0/#cccl_headers_have_moved_in_cuda_130) - MSCCLPP build fails due to stale include [here](https://github.com/microsoft/mscclpp/blob/960a8ddebde7e4489130f624524397fef6bdda38/include/mscclpp/atomic_device.hpp#L10C1-L10C23) - Quickest fix for me was to change this [blob](https://github.com/microsoft/mscclpp/blob/960a8ddebde7e4489130f624524397fef6bdda38/CMakeLists.txt#L125C1-L134C7) from `CMakeLists.txt` as below. ```CMake if(MSCCLPP_USE_CUDA) #... find_package(CCCL PATHS ${CUDAToolkit_LIBRARY_DIR}/cmake/cccl) if (CCCL_FOUND AND CUDAToolkit_VERSION_MAJOR GREATER 12) set(GPU_LIBRARIES CUDA::cudart CUDA::cuda_driver CCCL::CCCL) elseif() set(GPU_LIBRARIES CUDA::cudart CUDA::cuda_driver) endif() #... else() ```