leezu commented on issue #17293: [Build] Add a reasonable default for CMAKE_CUDA_COMPILER in *nix URL: https://github.com/apache/incubator-mxnet/pull/17293#issuecomment-574643566 It's expected that to compile software, the compiler must be available. For that, it must be either on `$PATH` or manually specified. With respect to C and C++ compilers, this is what the CC and CXX environment variables are for. For example, `CC=gcc-9 CXX=g++-9 cmake ..` will prepare the build with GCC 9. Alike, if users want to use non-standard nvcc (ie nvcc that is not on PATH), they can set CUDACXX. `CUDACXX=/usr/local/cuda/nvcc cmake ..`. I think it's better to follow standard practice instead of taking additional assumptions. For example, we may want to use clang to compile the cuda files instead of nvcc. If nvcc is not on path, users may reasonably expect that clang will be used. What do you think?
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
