线程数增加，差距继续拉大。
(test314) PS Z:\py> python3.14t.exe .\ng.py 5
5 线程(noGIL) 耗时 0.64s  结果 170666663466666680000000
(test314) PS Z:\py> python3.14t.exe .\ng.py 10
10 线程(noGIL) 耗时 0.41s  结果 170666663466666680000000
(test314) PS Z:\py> python3.14t.exe .\ng.py 15
15 线程(noGIL) 耗时 0.34s  结果 170666663466666680000000
有GIL的版本
(test314) PS Z:\py> python3.14.exe .\ng.py 15
15 线程(noGIL) 耗时 2.78s  结果 170666663466666680000000

# 需要 Python 3.14 free-threaded
"""
ng.py  [M]

在 Python 3.14 free-threaded（--disable-gil）下，
用 M 个线程并行计算 0…N-1 的平方和。
M 省略时默认=CPU 核数；若给出，则必须是 1<=M<500 的正整数。
"""
import os, sys, threading, time
from concurrent.futures import ThreadPoolExecutor

N = 80_000_000          # 任务总量
DEFAULT_M = os.cpu_count()

def partial_sum(start: int, end: int) -> int:
    s = 0
    for i in range(start, end):
        s += i * i
    return s

def parse_workers() -> int:
    if len(sys.argv) == 1:
        return DEFAULT_M
    try:
        m = int(sys.argv[1])
        if 1 <= m < 500:
            return m
        raise ValueError
    except ValueError:
        print("用法: python3.14t ng.py  [M]\n"
              "可选参数 M 必须是 1 到 499 之间的整数", file=sys.stderr)
        sys.exit(1)

def run(workers: int) -> int:
    chunk = N // workers
    ranges = [(i * chunk, (i + 1) * chunk if i != workers - 1 else N)
              for i in range(workers)]

    ta = time.perf_counter()
    with ThreadPoolExecutor(max_workers=workers) as pool:
        futures = [pool.submit(partial_sum, st, ed) for st, ed in ranges]
        total = sum(f.result() for f in futures)
    tb = time.perf_counter()

    print(f"{workers} 线程(noGIL) 耗时 {tb-ta:.2f}s  结果 {total}")
    return total

if __name__ == "__main__":
    run(parse_workers())

※ 修改:·poocp 于 Oct 20 06:11:00 2025 修改本文·[FROM: 171.221.52.*]
※ 来源:·水木社区 http://www.mysmth.net·[FROM: 171.221.52.*]

修改:poocp FROM 171.221.52.*
FROM 171.221.52.*