GCD 管理最大并发数

先前,项目中, 当时出现一些摸不着头脑的崩溃, 初步分析是 线程管理没限制好 导致的.
并且当时记录有Demo, 一直还未分享, 正好今夜打完🏀 精力旺盛有time ,挑灯夜笔挥一把~
bugly 问题现象如下

image.png

特此, 把各种线程管理方式 收集 罗列一下

---

< 0 >. GCD没有像NSOperation提供MaxConcurrentOperationCount方法, 所以一般我们在偶尔一次异步时这么用

// GCD, dispatch_async
- (void)test_GCD_dispatch_async {

    NSLog(@"begin: %@",[NSThread currentThread]);
    for (int i = 0; i < 10; i++) {
        __block int index = i;
        
        dispatch_async(dispatch_queue_create("socket msg recv queue", NULL), ^{
        
            sleep(1);
            NSLog(@"执行第%zd次操作，线程：%@",index, [NSThread currentThread]);
        
        });
    }
// count:   10
// begin:   02:52:54.213
// end  :   02:52:55.231
// time :    ≈ 1s
}

但是像上面那么用, 当用在接收socket消息时, 由于消息收的很快,短时间会开N个线程,而且在里面处理耗时任务时, 没等到达到60多个线程的上限, 可能主线程资源就被挤掉, 造成开篇所示的现象了 crash or stutter
【当然，用GCDSocket时，你只需传一个异步串行队列进去，内部在消息回调时只会开一个异步线程，不会n个n个的开了】
所以, 用时需谨慎。 那么,怎么实现线程数目的控制,这里有答案👇

---

< 1 >. 当然, 多线程管理, 首选NSOperation, SD、AFN等都在用, 它是对GCD的一层封装,方便线程的管理

//  NSOperationQueue
- (void)test_NSOperationQueue_withMaxTheadCount:(NSInteger) max {
    
    NSOperationQueue *queue = [[NSOperationQueue alloc] init];
    
    [queue setMaxConcurrentOperationCount:max];
    
    NSLog(@"begin: %@",[NSThread currentThread]);
    for (int i = 0; i < 10; i++) {
        
        __block int index = i;
        NSBlockOperation *operation = [NSBlockOperation blockOperationWithBlock:^(){
            sleep(1);
            NSLog(@"执行第%zd次操作，线程：%@",index, [NSThread currentThread]);
        }];
        
        [queue addOperation:operation];

    }
    
// max  :   2               10
// begin:   02:29:49.548    02:41:20.911
// end  :   02:29:54.806    02:41:21.985
// time :      ≈ 5.26s         ≈ 1s
}

< 2 >. 通过信号量semaphore控制

// GCD, semaphore
- (void)test_GCD_semaphore_WithMaxTheadCount:(NSInteger) max {
    
    dispatch_queue_t workConcurrentQueue = dispatch_queue_create("cccccccc", DISPATCH_QUEUE_CONCURRENT);
    dispatch_queue_t serialQueue = dispatch_queue_create("sssssssss",DISPATCH_QUEUE_SERIAL);
    dispatch_semaphore_t semaphore = dispatch_semaphore_create(max);    // create semaphore: value = 3

    NSLog(@"begin: %@",[NSThread currentThread]);
    for (int i = 0; i < 10; i++) {
        __block int index = i;
        dispatch_async(serialQueue, ^{
            
            // If value < 0, then wait here. Else value > 0, then pass, and value -1
            dispatch_semaphore_wait(semaphore, DISPATCH_TIME_FOREVER);
            dispatch_async(workConcurrentQueue, ^{
                
                sleep(1);
                NSLog(@"执行第%zd次操作，线程：%@",index, [NSThread currentThread]);
                dispatch_semaphore_signal(semaphore);});                // Perform value +1
        });
    }
    NSLog(@"主线程...!");
// max  :   2               10
// begin:   03:02:45.309    03:05:53.301
// end  :   03:02:50.332    03:05:54.309
// time :    ≈ 5.03s           ≈ 1s
}

< 3 >. 根据<2> 封装成 QSDispatchQueue, 使用如下👇

// QSDispatchQueue
- (void)test_QSDispatchQueue_withMaxTheadCount:(NSInteger) max {

    dispatch_queue_t concurrentQueue = dispatch_queue_create("socketConcurrentQueue", DISPATCH_QUEUE_CONCURRENT);
    QSDispatchQueue *queue = [[QSDispatchQueue alloc]initWithQueue:concurrentQueue concurrentCount:max];
    
    NSLog(@"begin: %@",[NSThread currentThread]);
    for (int i = 0; i < 10; i++) {
        __block int index = i;
        [queue async:^{
            sleep(1);
            NSLog(@"执行第%zd次操作，线程：%@",index, [NSThread currentThread]);
        }];
    }
    
// max  :   2               10
// begin:   02:37:29.052    02:40:11.811
// end  :   02:37:34.069    02:40:12.816
// time :    ≈ 5.1s           ≈ 1s
}

< 2 >. 通过dispatch_group_t控制

dispatch_group_t group = dispatch_group_create();

dispatch_group_enter(group);
[self request1:^{
        dispatch_group_leave(group);
    }];

dispatch_group_enter(group);
    [self request2:^{
        dispatch_group_leave(group);
    }];
    
dispatch_group_notify(group, dispatch_get_main_queue(), ^{
    //1、2两请求完毕
});

懒人下载 👉GitHub Demo