Compiling CUDA programs with clang takes too long.

[yanan-sjh]

I am using clang-18 to compile CUDA programs, and the compilation process does not report any errors, but it takes a very long time (even over an hour).

[yanan-sjh]

clang.zip

The attached files contain some examples that may be somewhat complex, but they can all be successfully compiled using nvcc in a very short amount of time.

[Artem-B]

clang++-18 -v host.cu kernel.cu -o clang.bin --cuda-gpu-arch=sm_89 -L/usr/local/cuda/lib64 -lcudart_static -ldl -lrt -pthread

I'd start with enabling optimizations and see if it helps. Unlike clang, nvcc has optimizations enabled by default, if I remember correctly.