Question

有没有办法证明“一次只能有一个线程执行代码”？

一个奇怪的发现是，我看到多个内核在htop中运行，好像有多个线程同时在执行，那是什么？

（我想这样做的原因是，我正在使用pybind将cpp代码绑定到python，并且正在处理一些GIL释放策略，所以我想看看有多少线程正在执行）

Answer 1

尝试运行一些使用大量CPU的难题，例如计算斐波纳契数列的第15800000个项。使用IDLE在单线程上花费大约2秒钟。现在就尝试两个。

import threading
import timeit

def tesr():
    a, b = 1, 0
    for _ in range(15800000):
        a, b = a + b, b

# Current thread
print("Current thread time:")
print(timeit.timeit(
    stmt='tesr()',
    setup='from __main__ import tesr',
    number=1,
))
print()

# Single thread
print("Single thread time:")
print("(Should take about the same time as current)")
t = threading.Thread(target=tesr)
print(timeit.timeit(
    stmt='t.start(); t.join()',
    setup='from __main__ import t',
    number=1,
))
print()

# Two threads
t1, t2 = (threading.Thread(target=tesr) for _ in range(2))
print("Two threads time:")
print("(Should take about double the current / single time)")
print(timeit.timeit(
    stmt='t1.start(); t2.start(); t1.join(); t2.join()',
    setup='from __main__ import t1, t2',
    number=1,
))

我的输出：

Current thread time:
2.0613602900000387

Single thread time:
(Should take about the same time as current)
2.228870080999968

Two threads time:
(Should take about double the current / single time)
4.671865998000044

如何演示python GIL（我怎么知道一次只执行一个线程）

1 个答案: