There is some function that indicates that one thread just can do what it have to do after another thread.
for example:
I have a vector like v = {1,2,3,4,5} and I have just 5 threads and I have to ensure that the thread 0 will be read the number 1, the thread 1 the number 2 and so on.
That’s a bit of a vague question - assuming you’re referring to device code, you can index into arrays based on the thread ID. As such, you can indeed access elements based on the thread number. Take a look at threadIdx and blockIdx (and the CUDA programming guide).
i.e. if you have a kernel launched with 1 thread block, consisting of threads in 1 dimension (such as having it declared as dim3(32,1,1)), then inside your kernel, threadIdx.x will give the thread dimension (along its x axis). Then if you have an array, you can use the thread ID to index into it.
Let me know if I misunderstood your question