我想创建一个 C 程序，它使用线程计算两个 N*N 矩阵的乘法。
我通过使用多个线程引用矩阵乘法开始此代码，但不是为结果矩阵的每个单元格创建 N * N 个线程，我想创建 N 个线程来同时执行乘法，其中结果矩阵的每一行将通过一个不同的线程。
到目前为止，我的代码如下所示：
#include <stdio.h>
#include <pthread.h>
#include <stdlib.h>
#include <string.h>
#define N 2
struct v {
int i; //Row
int j; //Column
};
int A[N][N] = {{ 1, 2 }, { 3, 4 }};//Matrix 1
int B[N][N] = {{ 2, 3 }, { 4, 5 }};//Matrix 2
int C[N][N]; //Resulting Matrix
static void *fnc(void *arg) {
struct v *data = (struct v *)arg;
int l;
for (l = 0; l < N; l++) {
int i = ((struct v *)data[l]).i; //Row No
int j = ((struct v *)data[l]).j; //Column No
int accumulator = 0;
int d, sum = 0;
for (d = 0; d < N; d++) {
sum = sum + A[i][d] * B[d][j];
}
C[i][j] = sum;
sum = 0;
}
return;
}
int main() {
pthread_t threads[N];
int i, k;
for (i = 0; i < N; i++) {
struct v *data[N];
for (k = 0; k < N; k++) {
data[k] = (struct v *)malloc(sizeof(struct v));
data[k].i = i; //assign the row of C for thread to calculate
data[k].j = k; //assign the column of C for thread to calculate
}
//In this example it creates 2 threads with passing data.
//Data consists of row and column pairs for each thread, that will be calcuting the pairs.
//Consider first iteration of this loop:
//Thread 1 is created and data consists of (0,0) and (0,1) which are the targeted calculation cells for thread 1.
//In the second iteration: Thread 2 will have (1,0) and (1,1) pairs in its data.
pthread_create(&threads[i], NULL, fnc, data);
}
for (i = 0; i < N; i++)
pthread_join(threads[i], NULL);
for (i = 0; i < N; i++)
for (k = 0; k < N; k++)
printf("%d\t", C[i][k]);
return 0;
}
当我编译时
gcc thread.c -lpthread -lrt
我收到此错误：
thread.c: In function ‘fnc’:
thread.c:24:24: error: cannot convert to a pointer type
int i=((struct v*)data[l]).i;//Row No
^
thread.c:24:35: error: request for member ‘i’ in something not a structure or union
int i=((struct v*)data[l]).i;//Row No
^
thread.c:25:24: error: cannot convert to a pointer type
int j=((struct v*)data[l]).j;//Column No
^
thread.c:25:35: error: request for member ‘j’ in something not a structure or union
int j=((struct v*)data[l]).j;//Column No
^
thread.c: In function ‘main’:
thread.c:48:20: error: request for member ‘i’ in something not a structure or union
data[k].i = i; //assign the row of C for thread to calculate
^
thread.c:49:20: error: request for member ‘j’ in something not a structure or union
data[k].j = k; //assign the column of C for thread to calculate
我相信代码背后的逻辑是正确的，这肯定看起来像一个简单的参数传递问题。我放了整个代码，因为如果我收到您对问题的评论以及对所需工作的可能更好的解决方案，我会很高兴。
小智 1
您好，我已经使用 3*3 线程实现了 3*3 矩阵乘法。代码可以修改为 N*N。代码是不言自明的。
*header*/
#include <pthread.h>
#include <stdio.h>
#include<stdlib.h>
/*macro*/
#define NUM_THREADS 9
/*I have defined the structure here to pass the multiple arguments
when creating threads*/
/*structure declaration*/
struct thread_data
{
int element1[3];//passing the element of one matrix
int element2[3];//passing the element of other matrix
int sum;//storing the result of multiplication
}thread_data_array[NUM_THREADS];
/**
* @brief finds matrix multiplication function
* param [in] : accepts the thread arguments
*
* @return void pointer
*/
void *multiplication(void *threadarg)
{
int i;
struct thread_data *my_data;
my_data = (struct thread_data *) threadarg;/*type cast the
void pointer to the struct thread*/
//pointer
my_data->sum=0;
for(i=0;i<3;i++){/*multiply and add each element
of 1st matrix to 2nd matrix*/
my_data->sum=my_data->sum
+((my_data->element1[i])*(my_data->element2[i]));
}
}
/*main*/
int main (int argc, char *argv[])
{
pthread_t threads[NUM_THREADS];
int rc;
int t;
int matrix_b[3][3]={{156,223,345},{456,5678,656},{712,811,922}};
//to store the matrix
int matrix_a[3][3]={{123,278,335},{437,547,656},{734,832,944}};
int ans[9];
int i,j,count,z,k,l,q;
t=i=j=z=count=k=q=l=0;
/*
suppose the matrices are
1 2 3 1 2 3
4 5 6 1 2 3
7 8 9 1 2 3
so to perform the multiplication
row 1 2 3 will be multiplied too column
111 222 333 respectively.
so 123 shold be stored in 3 spreate threads
and 111 222 333 should be stored in 3 seprate threads
123 111 should be in common thread
123 222 should be in common thread
123 333 should be in common thread
in this way below logic i have
writtten.
It is hard to understand at first glance..
so enjoy!!!!!!
*/
while(z<3){
for(count=0;count<3;count++){/*store the each row of
first matrix 3 times in 3 seprate threads*/
for(j=0;j<3;j++)
thread_data_array[t].element1[j]
= matrix_a[i][j];
t++;
}
/*store the each column of the second matrix in the spreate threads*/
for(k=0;k<3;k++){
for(l=0;l<3;l++)
thread_data_array[q].element2[l]
= matrix_b[l][k];
q++;
}
i++;
z++;
}
/*we have total 9 element in resultant matrix,so to create it we should
create 9 threads*/
for(t=0; t<NUM_THREADS; t++){
rc =pthread_create (&threads[t], NULL, multiplication,
(void *) &thread_data_array[t]);
/*we are passing the function pointer and
the pointer to structure as the argument,since we want to pass multiple argumen
ts we are passing the structure poi
nter as the arguments.*/
if (rc){//if rc is not 0,error is there
printf("ERROR; return code from () is %d\n", rc);
exit(-1);
}
}
/*
NAME
pthread_join - join with a terminated thread
SYNOPSIS
#include <pthread.h>
int pthread_join(pthread_t thread_here, void **retval);
Compile and link with -pthread.
DESCRIPTION
The pthread_join() function waits for the thread specified by thread_here
to terminate. If that thread has already terminated, then
pthread_join() returns immediately. The thread specified by thread
must be joinable.
*/
for (i = 0; i < 9; i++)/*to wait for all the threads to complete*/
pthread_join(threads[i], NULL);
printf("%d %d %d\n",thread_data_array[0].sum,thread_data_array[1].sum,thread_data_array[2].sum);
printf("%d %d %d\n",thread_data_array[3].sum,thread_data_array[4].sum,thread_data_array[5].sum);
printf("%d %d %d\n",thread_data_array[6].sum,thread_data_array[7].sum,thread_data_array[8].sum);
}