Parallelism inside core

how it is managed the sequential ‘for’ cycle inside the core?
In sequentiale mode or exist a logic parallelism?

Any sequential for loop inside a kernel remains sequential with regards to the particular thread/block/grid index.