To use /MD flag for host, should one use --cudart shared?

Accelerated Computing CUDA CUDA NVCC Compiler
1

If I use static cudart with -Xcompiler /MD it emits

LINK : warning LNK4098: defaultlib ‘LIBCMT’ conflicts with use of other libs

It tries to link to LIBCMT. Is it mandatory to use shared cudart with /MD?

1 Like