Pull math functions from the global namespace only when compiling cuda code with nvcc. When compiling with clang, we want to use the std namespace.
1 file changed