write a matrix multiplication program where the work of calculating the product
ID: 3667941 • Letter: W
Question
write a matrix multiplication program where the work of calculating the product matrix is distributed among multiple threads created using the pthread library. To this end, your program must do the following:
Define two global matrices A[X][Y] & B[Y][Z] to store the two matrices to be multiplied. X and Y must be set using a #define directive.
Define a global matrix C[X][Z] to store the product of matrices A and B. Once again, Z must be set using a #define directive.
Define a pthread function called pthread_multiply to perform the actual matrix multiplication operation (Note: Recall that a pthread function must have a very specific prototype). As a starting point, a regular (non-threaded) function for matrix multiplication, called multiply, has been provided below.
void multiply()
{
int row,col,inner;
for (row = 0; row < X; row++) {
for (col = 0; col < Z; col++) {
for (inner = 0; inner < Y; inner++) {
C[row][col] += A[row][inner] * B[inner][col];
}
}
}
}
The pthread_multiply function you write will essentially be a modification of the above multiply function. Instead of computing the entire product matrix C, the pthread_multiply function will only calculate a subset of rows of the product matrix C. Parameters must be passed to pthread_multiply to tell it which subset of rows to compute (Note: recall that parameter passing to a pthread function is somewhat different because of its specific prototype).
Define a main function and in this function, do the following:
Initialize the matrices A, B and C as follows: each element in A & B must be initialized to the sum of its row and column index and all elements in matrix C must be initialized to zero.
Using a loop, create multiple pthreads to execute the pthread_multiply function (use a #define directive to define the number of threads - this number must be set to more than 2). To each thread, pass appropriate parameters to tell it which subset of rows in the resultant matrix C it is responsible for computing. Each thread must be given an equal share of the resultant matrix as far as possible. For e.g., if you have A[10][10], B[10][10] and if the number of threads is 2 (Note: this is just an example - you must use more than 2 threads), then, thread 1 should be assigned the task of calculating results for rows 0 – 4 of matrix C and thread 2, those for rows 5 – 9. However, if you have A[9][9], B[9][9] and two threads, one of the threads will be assigned a lower number of rows than the other. I.e., in general, the last thread may have less number of rows to compute.
Using a second loop, wait until all the created pthreads have completed, using the pthread_join function.
Once all threads have completed, display the matrix C in an appropriate manner.
Explanation / Answer
# include <stdio.h>
# include <pthread.h>
# include <stdlib.h>
int M1[64][64];
int M2[64][64];
int M3[64][64];
void *pthread_multiply (void *);
struct coord {
int i; //Row
int j; //Column
};
int main()
{
int i,j,l,k=1,count=0;
//First Matrix
for (i=0;i<64;i++)
{
for (j=0;j<64;j++)
{
M1[i][j]=k;
k++;
}
}
//Second Matrix
i=0,j=0,k=1;
for (i=0;i<64;i++)
{
for (j=0;j<64;j++)
{
M2[i][j]=k;
k++;
}
}
//Multiplying the two Matrices
i=0,j=0,k=1;
int sum=0;
for (i=0;i<64;i++)
{
for (j=0;j<64;j++)
{
for (l=0;l<8;l++)
{
//Assign a row and column for each thread
struct coord *data = (struct coord *) malloc(sizeof(struct coord));
data->i = i;
data->j = j;
pthread_t tid; //Thread ID
pthread_attr_t attr; //Set of thread attributes
pthread_attr_init(&attr);
//Creating the thread
pthread_create(&tid,&attr,pthread_multiply,data);
//Parents wait for all thread to complete
pthread_join(tid, NULL);
}
}
}
printf(" Displaying multiplied matrices ");
i,j=0;
for (i=0;i<64;i++)
{
for(j=0;j<64;j++)
{
printf("%d",M3[i][j]);
printf(" ");
}
printf(" ");
}
return 0;
}
void *pthread_multiply(void *para)
{
struct coord *data = para;
int num, sum = 0;
//Row multiplied by column
for(num = 0; num< 64; num++)
{
sum += M1[data->i][num] * M2[num][data->j];
}
//assign the sum to its coordinate
M3[data->i][data->j] = sum;
//Exit the thread
pthread_exit(0);
}
Related Questions
Navigate
Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.