Python性能提升一 —— 高计算模块使用C++编写

场景

假设有一个高计算场景：双层循环，复杂度O(n*n)，并且n足够大。

纯python方式

def func(times):
    s = 0
    for i in range(times):
        for j in range(times):
            s += i*j
    return s

Python + C++库方式

高计算模块放到C++库中

long long TestFact::func(int n)
{
    long long lResult = 0;
    for (int i = 0; i < n; i++)
        for (int j = 0; j < n; j++)
            lResult += i*j;
    return lResult;
}

两种方式性能对比

附完整代码

C++库代码

#include <Python.h>

class TestFact{
public:
    TestFact(){};
    ~TestFact(){};
    long long func(int n);
};

long long TestFact::func(int n)
{
    long long lResult = 0;
    for (int i = 0; i < n; i++)
        for (int j = 0; j < n; j++)
            lResult += i*j;
    return lResult;
}

extern "C"
long long func(int n)
{
    TestFact t;
    return t.func(n);
}

编译成so库

g++ -fPIC -shared example.cpp -o example.so -I/usr/local/include/python2.6

Python代码

#!/usr/bin/python
#coding: utf-8
import os, time
import ctypes

def func(times):
    s = 0
    for i in range(times):
        for j in range(times):
            s += i*j
    return s

def main():
    n = 20000

    start = time.time()
    s = func(n)
    elapsed = time.time() - start
    print 'Pure Python: Result:%d, Time elapsed: %f s' % (s, elapsed)

    pdll = ctypes.CDLL('./example.so')
    pdll.func.restype = ctypes.c_longlong
    start = time.time()
    s = pdll.func(n)
    elapsed = time.time() - start
    print 'Python + C++: Result:%d, Time elapsed: %f s' % (s, elapsed)

if __name__ == '__main__':
    main()