multithreading - C++ matrix multiplication with multithreads and semaphore -

i have written program multiplication of 2 matrices. when execute it, works sometimes, not work , half of answers right. think there problem in usage of thread , semaphore together.

#include <iostream> #include <stdio.h> #include <stdlib.h> #include <mutex> #include <thread>  //note: code needs c++11 compiler compile.  using namespace std; int const size = 4; mutex mtx[size][size]; thread mythread[size];   void matrixmultiplication(int const * const * matrixa, int const * const * matrixb,int** matrixc, int row, int col, int i){     mtx[row][col].lock();     matrixc[row][col] += matrixa[row][i]*matrixb[i][col];     mtx[row][col].unlock(); }  int main() {     int** matrixa;     int** matrixb;     int** matrixc;      matrixa = new int*[size];     matrixb = new int*[size];     matrixc = new int*[size];      for(int = 0; i<size; i++){         matrixa[i] = new int[size];         matrixb[i] = new int[size];         matrixc[i] = new int[size];     }      //we initializing matrices.     int count = 1;     for(int = 0; i<size; i++){         for(int j = 0; j<size; j++){             matrixa[i][j] = count;             matrixb[i][j] = count;             count++;         }     }       for(int row=0; row<size; row++){         for(int col=0; col<size;col++){             for(int i=0;i<size;i++){                 mythread[i] = thread(matrixmultiplication, matrixa, matrixb, matrixc, row, col, i);                 mythread[i].join();             }         }     }       for(int = 0; i<size; i++){         for(int j = 0; j<size; j++){             cout<<matrixc[i][j]<<"  ";         }         cout<<endl;     }      return 0; } 

there 2 things going wrong here far can see.

   for(int row=0; row<size; row++){         for(int col=0; col<size;col++){             for(int i=0;i<size;i++){                 mythread[i] = thread(matrixmultiplication, matrixa, matrixb, matrixc, row, col, i);                 mythread[i].join();             }         }     } 

should be

   for(int row=0; row<size; row++){         for(int col=0; col<size;col++){             for(int i=0;i<size;i++){                 mythread[i] = thread(matrixmultiplication, matrixa, matrixb, matrixc, row, col, i);             }         }     }  for(int row=0; row<size; row++){    for(int col=0; col<size;col++){       for(int i=0;i<size;i++){          mythread[i].join();       }    } } 

this way first create bunch of calculations. , actual calculations. still not way should approach this. thread should have reasonable amount of work.

so before go ahead , that, allow me explain threading little bit. thread different work at same time. since doing @ same time, can never let thread access data might written thread. otherwise might end data that's half written to( called race condition).

therefore std::thread() call copy arguments given. if it's pointer. can let each thread calculations each column on separate thread. however, in order recommend creating column struct or class. , passing thread std::ref. passing std::ref prevent data being copied.