【Linux C 几种锁的性能对比】 1.读写锁 2.互斥锁 3.自旋锁 4.信号量 5.rcu

2023-12-30 10:45:21

直接上代码
rcu.c

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

#include <pthread.h>
#include <limits.h>
#include <semaphore.h>
#include <urcu.h>


/* 1.读写锁
   2.互斥锁
   3.自旋锁
   4.信号量
   5.rcu
*/


#define RW_LOCK  0
#define MUTEX_LOCK  0
#define SPIN_LOCK 0
#define _SEM  0
#define URCU  1 


struct point{
	int x;
	int y;
};

struct point *gp;
int done = 0;
long reads = 0;
pthread_rwlock_t rwlock;
pthread_mutex_t mutex_t;
pthread_spinlock_t spinlock;
sem_t sem;

void *timer(void *arg){
	struct timespec ts, ts2;
	timespec_get(&ts, TIME_UTC);

	while(!done){
		
		sleep(1);
		timespec_get(&ts2, TIME_UTC);

		time_t sec = ts2.tv_sec - ts.tv_sec;

		printf("reads: %ld, %ld K reads/sec\n", reads, (reads/sec)/1000);
	}
}

void *updater(void *arg){
	struct point *p;
	struct point *old;
	int i = 0;
	for(i = 0; i< INT_MAX; i ++){

		p = malloc(sizeof(struct point));

		p->x = i;
		p->y = i+1;

		old = gp;
#if 0
		gp = p;
#elif RW_LOCK
		pthread_rwlock_wrlock(&rwlock);
		gp = p;
		pthread_rwlock_unlock(&rwlock);
#elif MUTEX_LOCK
		pthread_mutex_lock(&mutex_t);
		gp = p;
		pthread_mutex_unlock(&mutex_t);
#elif SPIN_LOCK 
		pthread_spin_lock(&spinlock);
		gp = p;
		pthread_spin_unlock(&spinlock);
#elif URCU
		rcu_assign_pointer(gp, p);
		synchronize_rcu();
#else 
		sem_wait(&sem);
		gp = p;
		sem_post(&sem);
#endif
		free(old);
	}
}

void *reader(void *arg){

	rcu_register_thread();//urcu

	while(!done){
		int x, y;
#if 0
		x = gp->x;
		y = gp->y;
#elif RW_LOCK 
		pthread_rwlock_rdlock(&rwlock);
		x = gp->x;
		y = gp->y;
		pthread_rwlock_unlock(&rwlock);
#elif MUTEX_LOCK   
		pthread_mutex_lock(&mutex_t);
		x = gp->x;
		y = gp->y;
		pthread_mutex_unlock(&mutex_t);
#elif SPIN_LOCK 
		pthread_spin_lock(&spinlock);
		x = gp->x;
		y = gp->y;
		pthread_spin_unlock(&spinlock);
#elif URCU
		rcu_read_lock();
		struct point *p = rcu_dereference(gp);
		x = p->x;
		y = p->y;
		rcu_read_unlock();
#else
		sem_wait(&sem);
		x = gp->x;
		y = gp->y;
		sem_post(&sem);

#endif
		
		reads ++;
		if(y != x+1){
			printf("Error: x:%d, y:%d\n", x, y);
			done = 1;
			break;
		}
		
	}
	rcu_unregister_thread();
	exit(1);

}
// gcc -o rcu rcu.c -lpthread -lurcu
int main(){

	pthread_t tid[3];

	pthread_rwlock_init(&rwlock, NULL);
	pthread_mutex_init(&mutex_t, NULL);
	pthread_spin_init(&spinlock, PTHREAD_PROCESS_SHARED);
	sem_init(&sem, 0, 1);
	rcu_init();//rcu
	
	gp = malloc(sizeof(struct point));
	gp->x = 1;
	gp->y = 2;
	
	pthread_create(&tid[0], NULL, updater, NULL);
	pthread_create(&tid[1], NULL, reader, NULL);
	pthread_create(&tid[2], NULL, timer, NULL);

	int i = 0;
	for(i = 0; i < 3; i ++){
		pthread_join(tid[i], NULL);	
	}
	
	free(gp);
	pthread_rwlock_destroy(&rwlock);
	pthread_mutex_destroy(&mutex_t);
	pthread_spin_destroy(&spinlock);
	return 0;
}

读写线程相当情况下 读写速度 对比
1.读写锁
rw
2.互斥锁
mutex
3.自旋锁
spin
4.信号量
sem
5.rcu
rcu
可以看出rcu读写性能优异 , 多线程同步不建议使用读写锁,嫌麻烦可以直接用互斥锁

文章来源:https://blog.csdn.net/weixin_45397344/article/details/135209177
本文来自互联网用户投稿,该文观点仅代表作者本人,不代表本站立场。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。