Link error when using cooperative groups (already tried -rdc=true)

Accelerated Computing CUDA CUDA Programming and Performance
1

I’ve seen lots of threads on this topic. None of them help. Here’s my build error:

$ nvcc -arch=sm_70 -o grid_sync grid_sync.cu -rdc=true -lcudadevrt
nvlink error   : Undefined reference to 'cudaCGGetIntrinsicHandle' in '/tmp/tmpxft_00007a41_00000000-10_grid_sync.o'

And here’s my code:

#include <cooperative_groups.h>

__global__ void vectorAdd() {
    cooperative_groups::grid_group g = cooperative_groups::this_grid();
}

int main() {
    vectorAdd<<<2, 256>>>();
}

I’m running on Ubuntu 18.04.

$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2017 NVIDIA Corporation
Built on Fri_Nov__3_21:07:56_CDT_2017
Cuda compilation tools, release 9.1, V9.1.85
2

Upgrading to CUDA toolkit v11.0 fixed it.