Thread Scheduling
Thread scheduling is divided into user-level thread scheduling and core-level thread scheduling, each of which has different working methods and characteristics.
User-level thread scheduling
- Scheduling mechanism: The core is not directly responsible for the scheduling of user-level threads, but allocates the CPU to a process (such as process A) and allocates a time slice for it. During this time slice, the thread scheduler inside process A is responsible for selecting the specific thread to be executed (such as thread A1 or A2).
- Thread switching: Multiple threads of process A can be switched within the process, and the core does not interfere with the switching of threads. Only when the core decides to switch to another process will the process-level switch occur.
- Performance: User-level thread switching uses machine instructions and has a fast switching speed.
- Blocking situation: When a user-level thread is blocked due to I/O operations and other reasons, the entire process will also be suspended, causing other threads in the process to be unable to run.
Core-level thread scheduling
- Scheduling mechanism: The core is directly responsible for the scheduling of threads. The core can schedule and switch threads of different processes, so that threads of different processes can be switched.
- Thread switching: The core will perform a complete context switch when switching threads of different processes to ensure the accuracy of system management and scheduling.
- Performance: Core-level thread switching requires a complete context switch, so the switching speed is relatively slow.
- Blocking situation: In the core-level thread mode, when a thread is blocked due to waiting for I/O, it will not cause the entire process to be suspended, and other threads in the process can continue to run.
Comparison between user-level thread scheduling and core-level thread scheduling
| Features | User-level thread scheduling | Core-level thread scheduling |
|---|---|---|
| Scheduling speed | Fast switching speed, directly using machine instructions | Slow switching speed, requiring complete context switching |
| Blocking processing | Blocking of a single thread will suspend the entire process | Blocking of a single thread will not affect other threads in the process |
| Scheduling control | Responsible by the thread scheduler inside the process | Scheduling and switching are directly responsible by the core |
User-level thread scheduling is suitable for lightweight tasks with no blocking requirements, while core-level thread scheduling is suitable for scenarios that require fine-grained management and support for high concurrency.