Currently the computing tasks of TiDB belong to two different types of tasks: cop task and root task.
cop task is the computing task which is pushed down to the KV end for distributed execution; root task is the computing task for single point execution on the TiDB end.
Generally the input data of root task comes from cop task; when root task processes data, cop task of TiKV can processes data at the same time and waits for the pull of root task of TiDB. Therefore, cop tasks can be considered as executed concurrently with root task; but their data has an upstream and downstream relationship. During the execution process, they are executed concurrently during some time. For example, the first cop task is processing the data in [100, 200] and the second cop task is processing the data in [1, 100]. For details, see Understanding the TiDB Query Plan.