Python's GIL implemented in pure Python

https://rushter.com/blog/python-gil-thread-scheduling/

 

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103

import threading from types import SimpleNamespace   DEFAULT_INTERVAL = 0.05   gil_mutex = threading.RLock() gil_condition = threading.Condition(lock=gil_mutex) switch_condition = threading.Condition()   # dictionary-like object that supports dot (attribute) syntax gil = SimpleNamespace(     drop_request=False,     locked=True,     switch_number=0,     last_holder=None,     eval_breaker=True )       def drop_gil(thread_id):     if not gil.locked:         raise Exception("GIL is not locked")       gil_mutex.acquire()       gil.last_holder = thread_id     gil.locked = False       # Signals that the GIL is now available for acquiring to the first awaiting thread     gil_condition.notify()       gil_mutex.release()       # force switching     # Lock current thread so it will not immediately reacquire the GIL     # this ensures that another GIL-awaiting thread have a chance to get scheduled       if gil.drop_request:         switch_condition.acquire()         if gil.last_holder == thread_id:             gil.drop_request = False             switch_condition.wait()           switch_condition.release()     def take_gil(thread_id):     gil_mutex.acquire()       while gil.locked:         saved_switchnum = gil.switch_number           # Release the lock and wait for a signal from a GIL holding thread,         # set drop_request=True if the wait is timed out           timed_out = not gil_condition.wait(timeout=DEFAULT_INTERVAL)           if timed_out and gil.locked and gil.switch_number == saved_switchnum:             gil.drop_request = True       # lock for force switching     switch_condition.acquire()       # Now we hold the GIL     gil.locked = True       if gil.last_holder != thread_id:         gil.last_holder = thread_id         gil.switch_number += 1       # force switching, send signal to drop_gil     switch_condition.notify()     switch_condition.release()       if gil.drop_request:         gil.drop_request = False       gil_mutex.release()     def execution_loop(target_function, thread_id):     # Compile Python function down to bytecode and execute it in the while loop       bytecode = compile(target_function)       while True:           # drop_request indicates that one or more threads are awaiting for the GIL         if gil.drop_request:             # release the gil from the current thread             drop_gil(thread_id)               # immediately request the GIL for the current thread             # at this point the thread will be waiting for GIL and suspended until the function return             take_gil(thread_id)           # bytecode execution logic, executes one instruction at a time         instruction = bytecode.next_instruction()         if instruction is not None:             execute_opcode(instruction)         else:             return

　　

Note that this code will not run if you will try to execute it, because it's missing bytecode execution logic.

Some things to note

• Each thread executes its code in the separate execution_loop which is run by the real OS threads.
• When Python creates a thread it calls the take_gil function before entering the execution_loop.
• Basically, the job of the GIL is to pause the while loop for all threads except for a thread that currently owns the GIL. For example, if you have three threads, two of them will be suspended. Typically but not necessarily, only one Python thread can execute Python opcodes at a time, and the rest will be waiting a split second of time until the GIL will be switched to them.
• The C implementation can be found here and here.

A comment from the source code describes the algorithm as follows:

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

/*    Notes about the implementation:      - The GIL is just a boolean variable (locked) whose access is protected      by a mutex (gil_mutex), and whose changes are signalled by a condition      variable (gil_cond). gil_mutex is taken for short periods of time,      and therefore mostly uncontended.      - In the GIL-holding thread, the main loop (PyEval_EvalFrameEx) must be      able to release the GIL on demand by another thread. A volatile boolean      variable (gil_drop_request) is used for that purpose, which is checked      at every turn of the eval loop. That variable is set after a wait of      `interval` microseconds on `gil_cond` has timed out.         [Actually, another volatile boolean variable (eval_breaker) is used        which ORs several conditions into one. Volatile booleans are        sufficient as inter-thread signalling means since Python is run        on cache-coherent architectures only.]      - A thread wanting to take the GIL will first let pass a given amount of      time (`interval` microseconds) before setting gil_drop_request. This      encourages a defined switching period, but doesn't enforce it since      opcodes can take an arbitrary time to execute.        The `interval` value is available for the user to read and modify      using the Python API `sys.{get,set}switchinterval()`.      - When a thread releases the GIL and gil_drop_request is set, that thread      ensures that another GIL-awaiting thread gets scheduled.      It does so by waiting on a condition variable (switch_cond) until      the value of last_holder is changed to something else than its      own thread state pointer, indicating that another thread was able to      take the GIL.        This is meant to prohibit the latency-adverse behaviour on multi-core      machines where one thread would speculatively release the GIL, but still      run and end up being the first to re-acquire it, making the "timeslices"      much longer than expected.      (Note: this mechanism is enabled with FORCE_SWITCHING above) */