We'll now discuss two important concepts in GPU programming—thread synchronization and thread intercommunication. Sometimes, we need to ensure that every single thread has reached the same exact line in the code before we continue with any further computation; we call this thread synchronization. Synchronization works hand-in-hand with thread intercommunication, that is, different threads passing and reading input from each other; in this case, we'll usually want to make sure that all of the threads are aligned at the same step in computation before any data is passed around. We'll start here by learning about the CUDA __syncthreads device function, which is used for synchronizing a single block in a kernel.