与Python 2/3中的线程不同的结果

Question

Below is the code in Python 3. It always can get 100000. Why it is wrong? I think it should have different results.

import time, _thread

global count
count = 0
def test():
    global count

    for i in range(0, 10000):
        count += 1

for i in range(0, 10):
    _thread.start_new_thread(test, ())
time.sleep(5)
print(count)

Below is the code in Python 2. It always has different result (random).

import time, thread

global count
count = 0
def test():
    global count

    for i in range(0, 10000):
        count += 1

for i in range(0, 10):
    thread.start_new_thread(test, ())
time.sleep(5)
print count

Answer 1

CPython 2 allowed threads to switch after a certain number of byte codes had executed; CPython 3.2 changed to allow threads to switch after a certain amount of time has passed. Your test() executes plenty of byte codes, but consumes little time. On my box, under Python 3 the displayed result becomes unpredictable if I add, eg, this near the start:

import sys
sys.setswitchinterval(sys.getswitchinterval() / 10.0)

That is, allow threads to switch after 10 times less time (than the default) has elapsed.

Also note that _thread is strongly discouraged in Python 3: that's why the leading underscore was added. Using threading.Thread instead, like:

for i in range(10):
    t = threading.Thread(target=test)
    t.start()