GCD相关基础知识

我看世界

原创文章，转载请注明：转自：Try_Try_Try

1.获取Dispatch Queue 的两种方法

• 方法一: 通过GCD api 生成队列dispatch_queue_create

  // 串行队列
    dispatch_queue_t mySerialDispatchQueue = dispatch_queue_create("com.example.gcd.MySerialDispatchQueue", NULL);
    dispatch_async(mySerialDispatchQueue, ^{
    });
    dispatch_release(mySerialDispatchQueue);
    
   // 并行队列
    dispatch_queue_t myConcurrentDispatchQueue = dispatch_queue_create("com.example.gcd.MyConcurrentDispatchQueue", DISPATCH_QUEUE_CONCURRENT);
    dispatch_async(myConcurrentDispatchQueue, ^{
        NSLog(@"block on myConcurrentDispatchQueue");
    });
    dispatch_release(myConcurrentDispatchQueue);

• 方法二: 获取系统标准提供的 Dispatch Queue

  // 主队列   
    dispatch_queue_t mainDispatchQueue =   dispatch_get_main_queue();
    
  // 全局并发队列(4种优先级)
    dispatch_queue_t globalDispatchQueueHeigh = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_HIGH, 0);
    dispatch_queue_t globalDispatchQueueDefault = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0);
    dispatch_queue_t globalDispatchQueueLow = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_LOW, 0);
    dispatch_queue_t globalDispatchQueueBackground = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_BACKGROUND, 0);

2.延时操作

• 3s后,将block追加到主队列中进行执行;

  dispatch_after(dispatch_time(DISPATCH_TIME_NOW, (int64_t)(3 * NSEC_PER_SEC)), dispatch_get_main_queue(), ^{
        
        NSLog(@"waited at least 3 seconds");
    });

3.队列组

• 在一系列操作完成之后,再执行其他内容

    dispatch_queue_t queue = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0);
    
    dispatch_group_t group = dispatch_group_create();
    
    dispatch_group_async(group, queue, ^{
        NSLog(@"blk0");
    });
    dispatch_group_async(group, queue, ^{
        NSLog(@"blk1");
    });
    dispatch_group_async(group, queue, ^{
        NSLog(@"blk2");
    });
    
    // 推荐
    dispatch_group_notify(group, dispatch_get_main_queue(), ^{
        NSLog(@"done");
    });

  // 不推荐
    dispatch_time_t time = dispatch_time(DISPATCH_TIME_NOW, 1);
    long result = dispatch_group_wait(group, time);
    if (result == 0) { // group 的全部处理执行结束
    }else {
    }
//    dispatch_group_wait(group, DISPATCH_TIME_FOREVER);
    
    dispatch_release(group);

4.栅栏函数

• 用于读写操作引起的相关内容;
• 每次的执行结果可能不一样,但是写操作一定会将0,1,2,3读操作和4,5,6,7分隔开,从而达到不会误读的效果

    dispatch_queue_t myConcurrentDispatchQueue = dispatch_queue_create("com.example.gcd.MyConcurrentDispatchQueue", DISPATCH_QUEUE_CONCURRENT);
    
    __block NSInteger count = 1;
    
    void (^blk0_for_reading)(void) = ^{NSLog(@"blk0_for_reading-%li%@", count, [NSThread currentThread]);};
    void (^blk1_for_reading)(void) = ^{NSLog(@"blk1_for_reading-%li%@", count, [NSThread currentThread]);};
    void (^blk2_for_reading)(void) = ^{NSLog(@"blk2_for_reading-%li%@", count,[NSThread currentThread]);};
    void (^blk3_for_reading)(void) = ^{NSLog(@"blk3_for_reading-%li%@", count,[NSThread currentThread]);};
    
    void (^blk4_for_writing)(void) = ^{ ++count; NSLog(@"---------blk4_for_writing-%li%@", count,[NSThread currentThread]);};
    
    void (^blk5_for_reading)(void) = ^{NSLog(@"blk5_for_reading-%li%@", count,[NSThread currentThread]);};
    void (^blk6_for_reading)(void) = ^{NSLog(@"blk6_for_reading-%li%@", count,[NSThread currentThread]);};
    void (^blk7_for_reading)(void) = ^{NSLog(@"blk7_for_reading-%li%@", count,[NSThread currentThread]);};
    void (^blk8_for_reading)(void) = ^{NSLog(@"blk8_for_reading-%li%@", count,[NSThread currentThread]);};
    
    dispatch_async(myConcurrentDispatchQueue, blk0_for_reading);
    dispatch_async(myConcurrentDispatchQueue, blk1_for_reading);
    dispatch_async(myConcurrentDispatchQueue, blk2_for_reading);
    dispatch_async(myConcurrentDispatchQueue, blk3_for_reading);
    
    dispatch_barrier_async(myConcurrentDispatchQueue, blk4_for_writing);
    
    dispatch_async(myConcurrentDispatchQueue, blk5_for_reading);
    dispatch_async(myConcurrentDispatchQueue, blk6_for_reading);
    dispatch_async(myConcurrentDispatchQueue, blk7_for_reading);
    dispatch_async(myConcurrentDispatchQueue, blk8_for_reading);

// 结果
blk1_for_reading-1{number = 5, name = (null)}
2017-10-24 12:21:57.305030+0800 Test_copy_mutableCopy[1003:610433] blk0_for_reading-1{number = 4, name = (null)}
2017-10-24 12:21:57.305121+0800 Test_copy_mutableCopy[1003:610433] blk3_for_reading-1{number = 4, name = (null)}
2017-10-24 12:21:57.304973+0800 Test_copy_mutableCopy[1003:610435] blk2_for_reading-1{number = 5, name = (null)}
2017-10-24 12:21:57.305301+0800 Test_copy_mutableCopy[1003:610435] ---------blk4_for_writing-2{number = 5, name = (null)}
2017-10-24 12:21:57.305389+0800 Test_copy_mutableCopy[1003:610435] blk5_for_reading-2{number = 5, name = (null)}
2017-10-24 12:21:57.305460+0800 Test_copy_mutableCopy[1003:610435] blk6_for_reading-2{number = 5, name = (null)}
2017-10-24 12:21:57.305529+0800 Test_copy_mutableCopy[1003:610435] blk7_for_reading-2{number = 5, name = (null)}
2017-10-24 12:21:57.305598+0800 Test_copy_mutableCopy[1003:610433] blk8_for_reading-2{number = 4, name = (null)}

5.死锁的产生

必然引起死锁的两种方式

• 主队列 + 同步函数

    NSInteger count = 3;
    dispatch_queue_t mainQueue = dispatch_get_main_queue();
    dispatch_sync(mainQueue, ^{NSLog(@"mainQueue-%li%@", count,[NSThread currentThread]);});
  // result: 并没有出现结果,因为已经发生了死锁
  

• 串行队列 + 异步函数的block内部是(该串行队列 + 同步函数)(将串行队列换成主队列依然成立)

  dispatch_queue_t serialQueue = dispatch_queue_create("com.nextapp.qiannong", NULL);
    dispatch_async(serialQueue, ^{
        dispatch_sync(serialQueue, ^{NSLog(@"no deadLock");});
    });
  // result: 并没有出现结果,因为已经发生了死锁
  

6.dispatch_apply

• dispatch_apply 与dispatch_sync 函数类似(dispatch_group_async, dispatch_group_notify, dispatch_group_wait),会等待处理执行结束,在向下执行;

• dispatch_group_async 没有执行次数的要求

• dispatch_apply 必须有执行的次数

• dispatch_barrier_async ,用于前后都有读取异步任务,中间有一个写操作

  // 建议在dispatch_async中异步执行dispatch_apply函数
  
    dispatch_async(globalQueue, ^{
        dispatch_apply(10, globalQueue, ^(size_t index) {
            NSLog(@"%li-%@", index,[NSThread currentThread]);
        });
        dispatch_async(dispatch_get_main_queue(), ^{
            // 执行更新界面的操作
            NSLog(@"Done-%@", [NSThread currentThread]);
        });
    });
  
    // 试验之后,并没有生效
    dispatch_suspend(globalQueue); // 挂起
    dispatch_resume(globalQueue);  // 恢复
  

7.信号量机制相关

• 相比串行队列 和 栅栏函数 ,力度更小;

    dispatch_queue_t globalQueue = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0);
    
    // 保证可访问NSMutableArray类对象的线程只能有1个
    dispatch_semaphore_t semaphore = dispatch_semaphore_create(1);
    
    NSMutableArray *array = [NSMutableArray array];
    for (int idx = 0; idx < 10000; idx ++) {
        dispatch_async(globalQueue, ^{
 
           // 一直等到执行semaphore的是计数值>=1.
           //          long result = dispatch_semaphore_wait(semaphore, DISPATCH_TIME_FOREVER);
          //            if (result == 0) {
          //            }else {
          //            }
            
            [array addObject:[NSNumber numberWithInt:idx]];
            
            // 执行结束之后,释放该锁, 将dsemaphore的值 + 1;
            dispatch_semaphore_signal(semaphore);
        });
    }
    
    NSLog(@"%@", array);
    dispatch_release(semaphore);

8.单例

• 在程序中,只执行一次

  // 之前的初始化操作
    static BOOL initialized = NO;
    if (initialized == NO) {
        
        /*
         * 初始化操作
         */
        
        initialized = YES;
    }

  // 多线程环境下更加的安全
    static dispatch_once_t onceToken;
    dispatch_once(&onceToken, ^{
        
        /*
         * 初始化操作
         */
    });

9.简记

线程

• dispatch_sync/async:同步/异步是否开启新线程
• 异步不阻塞当前线程，同步阻塞当前线程
• serial/concurrent queue：串行/并行决定开启线程的条数

死锁

• 同步的在当前队列中(串行)中添加任务
• 主队列也是串行队列
• 相互阻塞构成死锁

信号量(semaphore)

• 信号量是同步的(同步会阻塞当前线程)
• semaphore.create(p): p >= 0
• semaphore.wait(): p <= 0, 阻塞当前线程， p -= 1
• semaphore.signal(): p += 1

group

• 实现多个网络的异步请求，最后再同步执行任务
• group.enter(), group.leave(), group.notify()