nvlink error when compiling CUDA code in linux

I am trying to run a dynamic parallelism algorithm using cuda. I compile my code as specified on the CUDA programming guide using the following line.

$ nvcc -arch=sm_35 -rdc=true kernel.cu -o output -lcudadevrt

My filename I want to compile is kernel.cu and it will output an executable.

However, I get the following two errors.
nvlink error : Undefined reference to ‘cudaGetParameterBufferV2’ in ‘/tmp/tmpxft_00001344_00000000-10_kernel.o’
nvlink error : Undefined reference to ‘cudaLaunchDeviceV2’ in ‘/tmp/tmpxft_00001344_00000000-10_kernel.o’

I am not sure if I am missing any dependencies or drivers or if this is a issue with my code itself. When I run my code without dynamic parallelism it compiles and executes as expected