多线程排序+快速排序

多线程排序，主要是将整个排序的序列分成若干份，每一个线程排序一份，所以线程排序完成之后，就进行归并，相当于多个有序序列合并成一个有序序列。

这里就需要用到线程屏障，也就是 pthread_barrier 系列函数。

屏障，通俗的说就是一个比赛跑步的过程，所以队员就绪了，才能进行比赛。

多线程排序也是，需要每个线程都是排序完成后，才能进行合并的过程。

代码：

#include <cstdio>#include <cstring>#include <iostream>#include <cstdlib>#include <sys/time.h>#include <pthread.h>#include <algorithm>using namespace std;const long MAX = 10000000L;  //数组中最大数const long long MAX_NUM = 100000000L;  //排序数const int thread = 4;       //线程数const int thread_num = MAX_NUM / thread;  //每个线程排序的个数int num[MAX_NUM];int tmp_num[MAX_NUM];pthread_barrier_t barrier;/** * Initialized Data */void init(){srandom((int)time(NULL));for(int i = 0; i < MAX_NUM; ++i)num[i] = random() % MAX;}/** *quick sort function */void qsorts(int *start, int *end){int nums = end - start;if(nums > 0){int index = 0;int flag = start[0];int i = 0, j = nums;while(i != j){while(j > i && start[j] >= flag)--j;start[index] = start[j];index = j;while(i < j && start[i] <= flag)++i;start[index] = start[i];index = i;}start[index] = flag;qsorts(start, start + (i - 1));qsorts(start + j + 1, end);}}void* work(void *arg)  //线程排序函数{long index = (long)arg;qsorts(num + index, num + index + thread_num - 1);pthread_barrier_wait(&barrier);pthread_exit(NULL);}void meger()        //最终合并函数{long index[thread];for (int i = 0; i < thread; ++i){index[i] = i * thread_num;}for(long i = 0; i < MAX_NUM; ++i){long min_index;long min_num = MAX;for(int j = 0; j < thread; ++j){if((index[j] < (j + 1) * thread_num) && (num[index[j]] < min_num)){min_index = j;min_num = num[index[j]];}}tmp_num[i] = num[index[min_index]];index[min_index]++;}}int main(int argc, char const *argv[]){init();struct timeval start, end;pthread_t ptid;//printf("%ld %ld\n", num[1], num[2]);gettimeofday(&start, NULL);//init pthread and Thread barrier//add 1, total have (thread + 1) threads.pthread_barrier_init(&barrier, NULL, thread + 1);for(int i = 0; i < thread; ++i)pthread_create(&ptid, NULL, work, (void *)(i * thread_num));pthread_barrier_wait(&barrier);meger();// use one thread to sort// qsorts(num, num + MAX_NUM - 1);gettimeofday(&end, NULL);long long s_usec = start.tv_sec * 1000000 + start.tv_usec;long long e_usec = end.tv_sec * 1000000 + end.tv_usec;double useTime = (double)(e_usec - s_usec) / 1000000.0;printf("sort use %.4f seconds\n", useTime);FILE *fp = fopen("result2.txt", "w+");for(long long i = 0; i < MAX_NUM; ++i)fprintf(fp, "%ld ", num[i]);return 0;}

当使用单线程排序时：

sort use 34.0476 seconds

当使用2线程排序时：

sort use 19.7711 seconds

当使用4线程排序时：

sort use 16.0659 seconds

当使用8线程排序时：

sort use 16.8172 seconds

如果使用STL中的函数的sort，将会更快。