C++ 多线程 矩阵相乘 C++ 11 thread 矩阵乘法快速实现

参考资料：https://zhuanlan.zhihu.com/p/146250334 （矩阵行列计算次序按照的是链接中提供的最优次序）

很惭愧地说，之前矩阵相乘一直在调函数，从来没有自己实现过，自己是非计算机科班出身，感觉基础确实一般。所以今天试着写了个多线程矩阵乘法作为练习，看看计算效率如何，如有问题欢迎交流指正。这里矩阵用的是 vector<vector<int> > 表示，绝对的速度上应该是不如数组，这里仅用于验证多线程的加速。

#include<iostream>
#include<vector>
#include<string>
#include<algorithm>
#include<thread>
#include<ctime>

using namespace std;

// f mn = d mt * e tn
vector<vector<int> > d;
vector<vector<int> > e;
vector<vector<int> > f;

//single thread
void mulSingle()
{
	int m = d.size();
	int t = d[0].size();
	int n = e[0].size();
	// ikj is faster than ijk (memory access issues)
	for (int i = 0; i < m; ++i)
		for (int k = 0; k < t; ++k)
		{
			int s = d[i][k];
			for (int j = 0; j < n; ++j)
				f[i][j] += s * e[k][j];
		}
}

//multi thread
void mulMulti(int rowStart, int rowEnd)
{
	int m = d.size();
	int t = d[0].size();
	int n = e[0].size();
	// ikj is faster than ijk (memory access issues)
	for (int i = rowStart; i < rowEnd; ++i)
		for (int k = 0; k < t; ++k)
		{
			int s = d[i][k];
			for (int j = 0; j < n; ++j)
				f[i][j] += s * e[k][j];
		}
}
// create a matrix
vector<vector<int> > createMat(int m, int n) {
	vector<vector<int> > ans(m, vector<int>(n, 0));
	for (int i = 0; i < m; i++)
		for (int j = 0; j < n; j++)
			ans[i][j] = i - j - 1;   // arbitrary value
	return ans;
}
int main()
{
	clock_t startTime, endTime;

	// initializing matrices
	d = createMat(80, 50);
	e = createMat(50, 22);

	// f (m*n) = d (m*t) * e (t*n)
	int m = d.size();
	int t = d[0].size();
	int n = e[0].size();

	f.resize(m);
	for (int i = 0; i < m; ++i)
		f[i].resize(n);

	//single thread
	startTime = clock();
	mulSingle();
	endTime = clock();

	//display
	for (auto cur : f)
	{
		for (auto i : cur)
			cout << i << " ";
		cout << endl;
	}	
	cout << "Single Thread Total Time : " << (double)(endTime - startTime)\
		/ CLOCKS_PER_SEC << " s" << endl;

	// initializing matrix
	f.clear();
	f.resize(m);
	for (int i = 0; i < m; ++i)
		f[i].resize(n);
	cout << endl;

	//multiple thread
	startTime = clock();
	int div = m / 4;
	thread t1(mulMulti, 0, div);
	thread t2(mulMulti, div, 2 * div);
	thread t3(mulMulti, 2 * div, 3 * div);
	thread t4(mulMulti, 3 * div, m);
	t1.join();
	t2.join();
	t3.join();
	t4.join();
	endTime = clock();

	//display
	for (auto cur : f)
	{
		for (auto i : cur)
			cout << i << " ";
		cout << endl;
	}	
	cout << "Multiple Thread Total Time : " << (double)(endTime - startTime)\
		/ CLOCKS_PER_SEC << " s" << endl;

	//this_thread::sleep_for(chrono::seconds(5));
	system("pause");
	return 0;
}

下图是单线程运算时间：0.063s

下图是多线程运算时间：0.041s

把矩阵行列数设大一些效果更明显