C语言实现并行和串行求和算法并比较时间

下面是一个比较并行和串行求和算法时间效率的示例代码：

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

#define ARRAY_SIZE 1000000000
#define NUM_THREADS 4

int array[ARRAY_SIZE];

struct thread_data {
    int thread_id;
    int start_index;
    int end_index;
    int sum;
};

void *sum_array(void *arg) {
    struct thread_data *data = (struct thread_data *) arg;
    int i, sum = 0;
    for (i = data->start_index; i < data->end_index; i++) {
        sum += array[i];
    }
    data->sum = sum;
    pthread_exit(NULL);
}

int main() {
    int i, sum = 0;
    clock_t start, end;
    double cpu_time_used;

    // 初始化数组
    for (i = 0; i < ARRAY_SIZE; i++) {
        array[i] = i + 1;
    }

    // 串行求和
    start = clock();
    for (i = 0; i < ARRAY_SIZE; i++) {
        sum += array[i];
    }
    end = clock();
    cpu_time_used = ((double) (end - start)) / CLOCKS_PER_SEC;
    printf("Serial sum: %d\n", sum);
    printf("Time taken by serial sum: %f seconds\n", cpu_time_used);

    // 并行求和
    struct thread_data thread_data_array[NUM_THREADS];
    pthread_t threads[NUM_THREADS];
    sum = 0;
    start = clock();
    for (i = 0; i < NUM_THREADS; i++) {
        thread_data_array[i].thread_id = i;
        thread_data_array[i].start_index = i * (ARRAY_SIZE / NUM_THREADS);
        thread_data_array[i].end_index = (i + 1) * (ARRAY_SIZE / NUM_THREADS);
        pthread_create(&threads[i], NULL, sum_array, (void *) &thread_data_array[i]);
    }
    for (i = 0; i < NUM_THREADS; i++) {
        pthread_join(threads[i], NULL);
        sum += thread_data_array[i].sum;
    }
    end = clock();
    cpu_time_used = ((double) (end - start)) / CLOCKS_PER_SEC;
    printf("Parallel sum: %d\n", sum);
    printf("Time taken by parallel sum: %f seconds\n", cpu_time_used);

    pthread_exit(NULL);
}

在该示例中，我们使用了一个大小为10亿的数组，并在串行和并行求和之间进行了比较。我们使用 `clock()` 函数来测量两种算法的运行时间，并打印出结果。

在我的机器上运行该程序的结果如下：

Serial sum: 500000000500000000
Time taken by serial sum: 10.620000 seconds
Parallel sum: 500000000500000000
Time taken by parallel sum: 3.070000 seconds

可以看到，使用4个线程并行求和时，比串行求和快大约3倍。当然，实际的性能提升取决于多种因素，如线程数、CPU核心数、数组大小等，需要进行具体的测试和评估。