C++多线程编程（第四章 案例1，C++11和C++17 多核并行计算样例）

4.1手动实现多核base16编码

4.1.1 实现base16编码

二进制转换为字符串
一个字节8位，拆分为两个4位字节（最大值16）
拆分后的字节映射到0123456789abcdef

4.1.2无多线程代码

#include <thread>
#include<iostream>
#include<future>
#include <string>
#include <vector>
#include <chrono>//计时头，C++11using namespace std;
using namespace chrono;static const char base16[] = "0123456789abcdef";
void Base16Encode(const unsigned char* data, int size, unsigned char* out)
{for (int i = 0; i < size; i++){unsigned char d = data[i];//0000 0000//1234 5678 >> 4              => 0000 1234 移位操作//1234 5678 & 0000 1111       => 0000 5678char a = base16[d >> 4];//取出高位char b = base16[d & 0x0F];//取出低位out[i * 2] = a;out[i * 2+1] = b;}
}int main()
{//使用字符串进行16位编码测试string test_data = "测试base16编码";printf("test_data size=%d\n", test_data.size());//一个中文占用2个字节，一个英文占用1个字节，如果要解码出来包含中文和英文，有点难度unsigned char* out = new unsigned char[test_data.size()*2+1];Base16Encode((unsigned char*)test_data.data(), test_data.size(), out);//编码out[test_data.size()*2] = '\0';//字符串\0结尾，否则后面总有乱码cout << "base16:" << out << endl;//测试单线程base16编码效率{vector<unsigned char>in_data;in_data.resize(1024 * 1024 * 20);//  1024*1024*10  -> 10M 测试数据for (int i = 0; i < in_data.size(); i++){//初始化里面数据in_data[i] = i % 256;}vector<unsigned char>out_data;out_data.resize(in_data.size()*2);//是输入数据2倍auto start = system_clock::now();//计时开始Base16Encode(in_data.data(), in_data.size(),out_data.data());auto end = system_clock::now();//计时结束auto duration = duration_cast<milliseconds>(end-start);//模板函数，转换到毫秒cout << "编码：" << in_data.size() << "字节数据花费" << duration.count() << "毫秒" << endl;//cout << out_data.data() << endl;}printf("All done!\n");getchar();return 0;
}

4.1.3 C++ 11多线程代码

//C++11多线程
void Base16EncodeThread(const vector<unsigned char>&data, vector<unsigned char> &out)
{long size = data.size();int th_count = thread::hardware_concurrency();//系统支持的线程核心数printf("CPU report thread count:%d\n",th_count);//对原始数据进行切片long slice_count = size / th_count;//余数丢弃了，余数后续单独处理if (size < th_count){//只切一片th_count = 1;//1个线程搞定slice_count = size;}vector<thread> ths;//准备好线程ths.resize(th_count);//任务分配到各个线程   for (int i = 0; i < th_count; i++){//eg. 1234 5678 9abc defg hi//计算偏移位置long offset = i * slice_count;long count = slice_count;//最后一个线程要把余数加起来一起处理if (th_count > 1 && i == th_count - 1){count = slice_count + size%th_count;}//cout << offset << ":" << count << endl;ths[i] = thread(Base16Encode, data.data() + offset,count, out.data());}//等待所有线程处理结束for (auto &th : ths){th.join();}
}

4.1.4 C++ 17多线程并发

		printf("C++17多线程Base16编码效率测试(编译的时候先检查设置C++17) 开始计算===========================\n");//设置C++17方法：属性->C/C++ ->C++语言标准  ，设置ISOC++17vector<unsigned char>in_data;in_data.resize(TestNumber);//  1024*1024*10  -> 10M 测试数据for (int i = 0; i < in_data.size(); i++){//初始化里面数据in_data[i] = i % 256;}vector<unsigned char>out_data;out_data.resize(in_data.size() * 2);//是输入数据2倍auto start = system_clock::now();//计时开始//#include <execution> //C++17 支持std::for_each(std::execution::par,//并行计算 多核in_data.begin(),in_data.end(),[&](auto& d)//多线程进入此函数{char a = base16[(d >> 4)];char b = base16[(d & 0x0F)];int index = &d - in_data.data();out_data[index * 2] = a;out_data[index * 2 + 1] = b;});auto end = system_clock::now();//计时结束auto duration = duration_cast<milliseconds>(end - start);//模板函数，转换到毫秒cout << "C++17多线程 编码：" << in_data.size() << "字节数据花费" << duration.count() << "毫秒" << endl;

4.1.5 所有测试代码汇总

#include <thread>
#include<iostream>
#include<future>
#include <string>
#include <vector>
#include <chrono>//计时头，C++11#include <execution> //C++17 for_eachusing namespace std;
using namespace chrono;static const char base16[] = "0123456789abcdef";
void Base16Encode(const unsigned char* data, long size, unsigned char* out)
{for (int i = 0; i < size; i++){unsigned char d = data[i];//0000 0000//1234 5678 >> 4              => 0000 1234 移位操作//1234 5678 & 0000 1111       => 0000 5678char a = base16[d >> 4];//取出高位char b = base16[d & 0x0F];//取出低位out[i * 2] = a;out[i * 2+1] = b;}
}//C++11多线程
void Base16EncodeThread(const vector<unsigned char>&data, vector<unsigned char> &out)
{long size = data.size();int th_count = thread::hardware_concurrency();//系统支持的线程核心数printf("CPU report thread count:%d\n",th_count);//对原始数据进行切片long slice_count = size / th_count;//余数丢弃了，余数后续单独处理if (size < th_count){//只切一片th_count = 1;//1个线程搞定slice_count = size;}vector<thread> ths;//准备好线程ths.resize(th_count);//任务分配到各个线程   for (int i = 0; i < th_count; i++){//eg. 1234 5678 9abc defg hi//计算偏移位置long offset = i * slice_count;long count = slice_count;//最后一个线程要把余数加起来一起处理if (th_count > 1 && i == th_count - 1){count = slice_count + size%th_count;}//cout << offset << ":" << count << endl;ths[i] = thread(Base16Encode, data.data() + offset,count, out.data());}//等待所有线程处理结束for (auto &th : ths){th.join();}
}int main()
{int TestNumber = 1024 * 1024 * 20 - 1; //1024 * 1024 * 10  -> 10M 测试数据大小//使用字符串进行16位编码测试string test_data = "测试base16编码";printf("test_data size=%d\n", test_data.size());//一个中文占用2个字节，一个英文占用1个字节，如果要解码出来包含中文和英文，有点难度unsigned char* out = new unsigned char[test_data.size()*2+1];Base16Encode((unsigned char*)test_data.data(), test_data.size(), out);//编码out[test_data.size()*2] = '\0';cout << "base16:" << out << endl;//测试单线程base16编码效率{printf("单线程Base16编码效率测试 开始计算===========================\n");vector<unsigned char>in_data;in_data.resize(TestNumber);//  1024*1024*10  -> 10M 测试数据for (int i = 0; i < in_data.size(); i++){//初始化里面数据in_data[i] = i % 256;}vector<unsigned char>out_data;out_data.resize(in_data.size()*2);//是输入数据2倍auto start = system_clock::now();//计时开始Base16Encode(in_data.data(), in_data.size(),out_data.data());auto end = system_clock::now();//计时结束auto duration = duration_cast<milliseconds>(end-start);//模板函数，转换到毫秒cout << "编码：" << in_data.size() << "字节数据花费" << duration.count() << "毫秒" << endl;//cout << out_data.data() << endl;//预览编码后的文本}//测试C++11多线程base16编码效率{printf("C++11多线程Base16编码效率测试 开始计算===========================\n");vector<unsigned char>in_data;in_data.resize(TestNumber);//  1024*1024*10  -> 10M 测试数据for (int i = 0; i < in_data.size(); i++){//初始化里面数据in_data[i] = i % 256;}vector<unsigned char>out_data;out_data.resize(in_data.size() * 2);//是输入数据2倍auto start = system_clock::now();//计时开始Base16EncodeThread(in_data, out_data);//多线程auto end = system_clock::now();//计时结束auto duration = duration_cast<milliseconds>(end - start);//模板函数，转换到毫秒cout << "C++11多线程 编码：" << in_data.size() << "字节数据花费" << duration.count() << "毫秒" << endl;//cout << out_data.data() << endl;//预览编码后的文本}//测试C++11多线程base16编码效率{printf("C++17多线程Base16编码效率测试(编译的时候先检查设置C++17) 开始计算===========================\n");//设置C++17方法：属性->C/C++ ->C++语言标准  ，设置ISOC++17vector<unsigned char>in_data;in_data.resize(TestNumber);//  1024*1024*10  -> 10M 测试数据for (int i = 0; i < in_data.size(); i++){//初始化里面数据in_data[i] = i % 256;}vector<unsigned char>out_data;out_data.resize(in_data.size() * 2);//是输入数据2倍auto start = system_clock::now();//计时开始//#include <execution> //C++17 支持std::for_each(std::execution::par,//并行计算 多核in_data.begin(),in_data.end(),[&](auto& d)//多线程进入此函数{char a = base16[(d >> 4)];char b = base16[(d & 0x0F)];int index = &d - in_data.data();out_data[index * 2] = a;out_data[index * 2 + 1] = b;});auto end = system_clock::now();//计时结束auto duration = duration_cast<milliseconds>(end - start);//模板函数，转换到毫秒cout << "C++17多线程 编码：" << in_data.size() << "字节数据花费" << duration.count() << "毫秒" << endl;//cout << out_data.data() << endl;//预览编码后的文本}printf("All done!\n");getchar();return 0;
}

分析：
release版本优化的比较多，之所以C++17耗时较长原因是进入多线程次数远远大于C++11,C++11只进入了12次，而C++17采用lambda表达式函数，进入了TestNumber次