Closed tsologub closed 5 years ago
not relevant anymore. can't reproduce.
I reproduced this issue with a Jetson Xavier NX on Jetpack 4.4 Cuda 10.2 (L4T Base Image R32.4.4).
@vmarkovtsev one simplification to CMakeLists.txt could be to add Cuda as a language:
It is no longer necessary to use this module or call find_package(CUDA) for compiling CUDA code. Instead, list CUDA among the languages named in the top-level call to the project() command, or call the enable_language() command with CUDA. Then one can add CUDA (.cu) sources to programs directly in calls to add_library() and add_executable(). Source: https://cmake.org/cmake/help/latest/module/FindCUDA.html
I don't know for certain, but this also might fix a bug in locating:
/usr/local/cuda-10.2/targets/aarch64-linux/lib/stubs/libcurand.so
To resolve this otherwise, I had to add that stubs directory to my path, which is not what CUDA recommends doing (they only recommend adding the cuda bin
to your path):
# per CUDA quick start guide (libKMCUDA also needs targets on path to resolve libcurand): https://docs.nvidia.com/cuda/cuda-quick-start-guide/index.html
ENV PATH=/usr/local/cuda-10.2/bin:/usr/local/cuda-10.2/targets/aarch64-linux/lib/stubs${PATH:+:${PATH}}
ENV LD_LIBRARY_PATH=/usr/local/cuda-10.2/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
RUN CUDA_ARCH=$architecture \
CUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda-10.2 \
python3 -m pip install libKMCUDA
Hi. I am using: cuda - 10.0 cmake - 3.15.2
During installation on Linux with
pip install libKMCUDA
got the following stacktrace. It seems that cmake can't find cuda libraries. Right?