pthread实现多线程矩阵

下面是一个使用pthread实现多线程矩阵乘法的示例代码：

#include <stdio.h>
#include <stdlib.h>
#include <pthread.h>

#define MAX_SIZE 100

int A[MAX_SIZE][MAX_SIZE], B[MAX_SIZE][MAX_SIZE], C[MAX_SIZE][MAX_SIZE];

int N, M, K; // A[N][M], B[M][K], C[N][K]
int num_threads;

void* multiply(void* arg) {
    int thread_num = *(int*)arg;

    int start = thread_num * N / num_threads;
    int end = (thread_num + 1) * N / num_threads;

    for (int i = start; i < end; i++) {
        for (int j = 0; j < K; j++) {
            C[i][j] = 0;
            for (int k = 0; k < M; k++) {
                C[i][j] += A[i][k] * B[k][j];
            }
        }
    }

    pthread_exit(NULL);
}

int main(int argc, char* argv[]) {
    if (argc < 4) {
        printf("Usage: %s <N> <M> <K> [num_threads]\n", argv[0]);
        return 1;
    }

    N = atoi(argv[1]);
    M = atoi(argv[2]);
    K = atoi(argv[3]);

    if (argc >= 5) {
        num_threads = atoi(argv[4]);
    } else {
        num_threads = 1;
    }

    printf("Matrix A (%d x %d):\n", N, M);
    for (int i = 0; i < N; i++) {
        for (int j = 0; j < M; j++) {
            A[i][j] = rand() % 10;
            printf("%d ", A[i][j]);
        }
        printf("\n");
    }

    printf("Matrix B (%d x %d):\n", M, K);
    for (int i = 0; i < M; i++) {
        for (int j = 0; j < K; j++) {
            B[i][j] = rand() % 10;
            printf("%d ", B[i][j]);
        }
        printf("\n");
    }

    pthread_t* threads = (pthread_t*)malloc(num_threads * sizeof(pthread_t));
    int* thread_nums = (int*)malloc(num_threads * sizeof(int));

    for (int i = 0; i < num_threads; i++) {
        thread_nums[i] = i;
        pthread_create(&threads[i], NULL, multiply, &thread_nums[i]);
    }

    for (int i = 0; i < num_threads; i++) {
        pthread_join(threads[i], NULL);
    }

    printf("Matrix C (%d x %d):\n", N, K);
    for (int i = 0; i < N; i++) {
        for (int j = 0; j < K; j++) {
            printf("%d ", C[i][j]);
        }
        printf("\n");
    }

    free(threads);
    free(thread_nums);

    return 0;
}

这个代码会读入三个整数N、M和K，分别表示矩阵A、B和C的维度，然后随机生成A和B矩阵的值，最后使用多线程计算C矩阵的值，并输出结果。

代码中使用了pthread库来创建线程，并使用了pthread_create()和pthread_join()函数来创建和等待线程。具体来说，代码中会先创建num_threads个线程，并将线程号存储在thread_nums数组中。然后每个线程会计算C矩阵的一部分，具体的计算范围是根据线程号和矩阵维度计算得到的。最后主线程会等待所有线程计算完毕，并输出C矩阵的值。