std :: string处理短字符串的性能 [英] std::string performance for handling short strings

问题描述

Bjarne Stroustrup和其他专家说，在中处理短字符串时，C ++比C更快Bjarne Stroustrup的文章和我上一个问题

但在我的测试中， C ++ 比 C 慢了110％ >

g ++版本是4.4.6（在CentOS 6.3上运行）。这是因为g ++ 4.4.6具有较少的c ++ 11特性，例如 Rvalue Reference（move semantics）？

测试结果

输出 $ time a.out input_file 减去执行时间无调用 compose_X（）函数

• cpp版本：0.192秒


• C版本：0.091秒

源代码使用 -O2

编辑编译的
$ b

compose_cpp（）和 compose_p（）来自Bjarne's文章。他说， compose_cpp（）比 compose_p（）更好。

如果我测试错误，我该如何改善测试？

  #include< iostream> 
 #include< fstream> 
 
 #include< cstdlib> 
 #include< cstring> 
 
 std :: string compose_cpp（const std :: string& name，const std :: string& domain）
 {
 return name +'@'+ domain; 
} 
 
 char * compose_c（const char * name，const char * domain）
 {
 char * res =（char *）malloc + strlen（domain）+2）; 
 char * p = strcpy（res，name）; 
 
 p + = strlen（name）; 
 * p ='@' 
 strcpy（p + 1，domain）; 
 
 return res; 
} 
 
 int main（int argc，char * argv []）
 {
 std :: ifstream ifs; 
 ifs.open（argv [1]）; 
 
 std :: string email，domain; 
 
 while（ifs.good（））
 {
 ifs>>电子邮件; 
 ifs>>域; 
 
 // std :: string composed = compose_cpp（email，domain）; 
 
 char * composed = compose_c（email.c_str（），domain.c_str（））; 
免费（组成）; 
} 
 
 ifs.close（）; 
} 
  

输入文件

输入文件为1毫线长。每行小于20字节，随机生成。

  $ head -n 10 input.txt.1m 
 9742720 1981857.com 
 22504 4127435.com 
 342760 69167.com 
 53075 26710.com 
 3837481 1851920.com 
 98441 278536.com 
 4503887 9588108。 com 
 193947 90885.com 
 42603 8166125.com 
 3587671 2972​​96.com 
  

解决方案

我只是想在这里猜测，因为我没有要测试的数据文件。我认为你的结果可能不符合Stroustrup的期望，因为他在这里说：

是的，C ++版本，因为它不必

但是，我的理解是

code> libstdc ++ 对所有字符串使用动态内存（零长度字符串除外）。查看这个最近的答案关于 std :: string 对象在 libstdc ++ 中的小尺寸的问题： http://stackoverflow.com/a/27631366/12711

有可能你会有一个实现使用短字符串优化（如MSVC - 我不知道如果clang的libc ++使用或不是）更好的结果。

Bjarne Stroustrup and other experts said that C++ is faster than C for handling for short strings in Bjarne Stroustrup' article and my previous question

But in my test, C++ was about 110% slower than C.

g++ version is 4.4.6 (runs on CentOS 6.3). Is this because g++ 4.4.6 has less c++11 feature such as Rvalue Reference (move semantics)?

Test Result

output of $ time a.out input_file minus execution time of no calling compose_X() function

• cpp version : 0.192 sec
• C version : 0.091 sec

source code

compiled with -O2

Edit

compose_cpp() and compose_p() come from Bjarne's article. He said that compose_cpp() is fater than compose_p(). I want to check this fact with real test.

If I have tested wrong way, how can I improve test?

#include <iostream>
#include <fstream>

#include <cstdlib>
#include <cstring>

std::string compose_cpp(const std::string& name, const std::string& domain)
{
    return name + '@' + domain;
}

char* compose_c(const char* name, const char* domain)
{
    char* res = (char*) malloc(strlen(name)+strlen(domain)+2);
    char* p = strcpy(res,name);

    p += strlen(name);
    *p = '@';
    strcpy(p+1,domain);

    return res;
}

int main(int argc, char* argv[])
{
    std::ifstream ifs;
    ifs.open(argv[1]);

    std::string email, domain;

    while (ifs.good())
    {
        ifs >> email;
        ifs >> domain;

        // std::string composed = compose_cpp(email, domain);

        char* composed = compose_c(email.c_str(), domain.c_str());
        free(composed);
    }

    ifs.close();
}

input file

input file is 1 millon lines long. every line is less than 20 bytes, generated randomly.

$ head -n 10 input.txt.1m
9742720 1981857.com
22504 4127435.com
342760 69167.com
53075 26710.com
3837481 1851920.com
98441 278536.com
4503887 9588108.com
193947 90885.com
42603 8166125.com
3587671 297296.com

解决方案

I'm just going to put a guess here because I don't have the data file to test with. I think your results may not match Stroustrup's expectation because of what he says here:

Yes, the C++ version, because it does not have to count the argument characters and does not use the free store (dynamic memory) for short argument strings.

However, my understanding is that libstdc++ uses dynamic memory for all strings (except for zero length strings). See this recent SO answer to a question about the small size of the std::string object in libstdc++: http://stackoverflow.com/a/27631366/12711

It's possible you would have better results with an implementation that uses the short string optimization (like MSVC - I'm not sure if clang's libc++ uses it or not).

这篇关于std :: string处理短字符串的性能的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！