Depricated cudaDeviceSynchronize() in Dynamic Parallelism

Here is a recent post of a similar nature, that may help.