手机开发 2016-09-30
GCD即为Grand Central Dispatch的缩写,是一种主要用于异步处理任务的安全的高性能解决方案。
在此不对比其他异步处理技术,只记录GCD的使用及说明。
先记录一些必要的概念:线程,同步,异步,并行队列,串行队列
线程:程序中任务执行流的最小单元。一个应用程序,一般存在一个进程(拥有独立内存空间),而每个进程可以有多个线程,即多个任务执行流,类似于工厂车间的流水线。每个应用程序至少存在一个线程,即为主线程,作为唯一可以控制UI的线程。
同步:主要指多个任务在同一个线程上依次执行。
异步:主要指多个任务在不同线程上同时执行。
队列:C语言中的队列指先进先出(FIFO)的数据结构,与栈的先进后出(FILO)相反。在GCD中分DISPATCH_QUEUE_SERIAL(串行队列)和DISPATCH_QUEUE_CONCURRENT(并行队列)。
并行队列:主要指队列中多个任务可以同时执行。
串行队列:主要指队列中多个任务只能依次执行。
逐个记录重要的方法:
dispatch_queue_t dispatch_get_main_queue(void)
获取主线程关联的队列。主队列是串行队列,该队列中任务都将由主线程执行,UI相关的操作必须加入此队列。
dispatch_queue_t dispatch_get_global_queue(long identifier, unsigned long flags);
获取全局通用的并行队列。第一个参数一般指定优先级,有如下选项:
DISPATCH_QUEUE_PRIORITY_HIGH
DISPATCH_QUEUE_PRIORITY_DEFAULT
DISPATCH_QUEUE_PRIORITY_LOW
DISPATCH_QUEUE_PRIORITY_BACKGROUND
第二个参数,用于标记,一般传入0即可。
dispatch_queue_t dispatch_queue_create(const char *label, dispatch_queue_attr_t attr)
自定义创建一个派发队列。第一个参数作为描述,例如"newcustomqueue";第二个参数一般指定队列内任务是串行(DISPATCH_QUEUE_SERIAL)还是并行(DISPATCH_QUEUE_CONCURRENT)类型。
void dispatch_async(dispatch_queue_t queue, dispatch_block_t block);
将一个block加入队列,异步执行。
void dispatch_sync(dispatch_queue_t queue, dispatch_block_t block);
将一个block加入队列,同步执行。
使用和测试以上方法
定义了几个队列:
dispatch_queue_t mainQueue = dispatch_get_main_queue();
dispatch_queue_t globalQueue = dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0);
dispatch_queue_t customSerialQueue = dispatch_queue_create("customserialqueue", DISPATCH_QUEUE_SERIAL);
dispatch_queue_t customConcurrentQueue = dispatch_queue_create("customconcurrentqueue", DISPATCH_QUEUE_CONCURRENT);
使用示例一:
总结:
上述代码,在同一次运行中,执行了多次,其中两次截图如上。
1.主队列必定在主线程中执行
2.在主线程中,使用dispatch_async异步执行非主队列中任务,将产生新的子线程
使用示例二:
总结:
主线程卡死。该方法在block执行完之前不会return,而block也是在当前线程执行,所以互相等待,造成死锁。
在主线程中使用同步执行主队列任务,将死锁。
使用示例三:
总结:
调用dispatch_sync将在当前线程上依次执行添加的任务,不管是否在同一队列中,也不管队列类型。
使用示例五:
LOG(@"测试GCD dispatch_async混合dispatch_sync"); LOG(@"mainThread : %@", [NSThread mainThread]); dispatch_async(customSerialQueue, ^{ LOG(@"level 1 dispatch_async No.1 customSerialQueue : %@", [NSThread currentThread]); dispatch_async(customSerialQueue, ^{ LOG(@"level 2 dispatch_async No.1 customSerialQueue : %@", [NSThread currentThread]); }); dispatch_async(customSerialQueue, ^{ LOG(@"level 2 dispatch_async No.2 customSerialQueue : %@", [NSThread currentThread]); }); dispatch_async(customSerialQueue, ^{ LOG(@"level 2 dispatch_async No.3 customSerialQueue : %@", [NSThread currentThread]); }); //============================== dispatch_async(customConcurrentQueue, ^{ LOG(@"level 2 dispatch_async No.4 customConcurrentQueue : %@", [NSThread currentThread]); }); dispatch_async(customConcurrentQueue, ^{ LOG(@"level 2 dispatch_async No.5 customConcurrentQueue : %@", [NSThread currentThread]); }); dispatch_async(customConcurrentQueue, ^{ LOG(@"level 2 dispatch_async No.6 customConcurrentQueue : %@", [NSThread currentThread]); }); //============================== dispatch_sync(customConcurrentQueue, ^{ LOG(@"level 2 dispatch_sync No.7 customConcurrentQueue : %@", [NSThread currentThread]); }); dispatch_sync(customConcurrentQueue, ^{ LOG(@"level 2 dispatch_sync No.8 customConcurrentQueue : %@", [NSThread currentThread]); }); dispatch_sync(customConcurrentQueue, ^{ LOG(@"level 2 dispatch_sync No.9 customConcurrentQueue : %@", [NSThread currentThread]); }); }); dispatch_async(customConcurrentQueue, ^{ LOG(@"level 1 dispatch_async No.2 customConcurrentQueue : %@", [NSThread currentThread]); dispatch_sync(customSerialQueue, ^{ LOG(@"level 2 dispatch_sync No.10 customSerialQueue : %@", [NSThread currentThread]); }); dispatch_sync(customSerialQueue, ^{ LOG(@"level 2 dispatch_sync No.11 customSerialQueue : %@", [NSThread currentThread]); }); dispatch_sync(customSerialQueue, ^{ LOG(@"level 2 dispatch_sync No.12 customSerialQueue : %@", [NSThread currentThread]); }); //============================== dispatch_async(customConcurrentQueue, ^{ LOG(@"level 2 dispatch_async No.13 customConcurrentQueue : %@", [NSThread currentThread]); }); dispatch_async(customConcurrentQueue, ^{ LOG(@"level 2 dispatch_async No.14 customConcurrentQueue : %@", [NSThread currentThread]); }); dispatch_async(customConcurrentQueue, ^{ LOG(@"level 2 dispatch_async No.15 customConcurrentQueue : %@", [NSThread currentThread]); }); //============================== dispatch_async(customSerialQueue, ^{ LOG(@"level 2 dispatch_async No.16 customSerialQueue : %@", [NSThread currentThread]); }); dispatch_async(customSerialQueue, ^{ LOG(@"level 2 dispatch_async No.17 customSerialQueue : %@", [NSThread currentThread]); }); dispatch_async(customSerialQueue, ^{ LOG(@"level 2 dispatch_async No.18 customSerialQueue : %@", [NSThread currentThread]); }); });
总结:
1.串行队列customSerialQueue,只关注任务加入的先后顺序,不管是同步还是异步执行,总体顺序不变
2.多个同步执行的同一串行队列,如果加入了另一个执行队列,他们的执行线程相同
3.多个异步执行的同一串行队列,如果加入了另一个执行队列,他们的执行线程相同
4.综合上述,使用dispatch_sync,一定是在当前线程执行;使用dispatch_async,不一定产生新的子线程,如果在同一级下,执行串行队列,将使用已存在的同一线程。
5.多个嵌套的异步执行并行队列,可能使用已存在的闲置线程
void dispatch_barrier_async(dispatch_queue_t queue, dispatch_block_t block);
该方法与dispatch_async类似,将block加入队列异步执行。注意当轮到该block执行时候,唯一异步执行,执行完毕后才会执行其他任务。所以只有加入并行队列,该方法才有意义。特别适用于加锁操作。
void dispatch_after(dispatch_time_t when, dispatch_queue_t queue, dispatch_block_t block);
将block将入队列,延迟一定时间后异步执行。
使用示例:
dispatch_after(dispatch_time(DISPATCH_TIME_NOW, (int64_t)(5.0 * NSEC_PER_SEC)), dispatch_get_main_queue(), ^{
//code to be executed after a specified delay
});
表示在5秒以后,将block加入主队列异步执行。
void dispatch_apply(size_t iterations, dispatch_queue_t queue, void (^block)(size_t));
将block加入队列指定次数。队列可以是并行的,所以block执行可以为并行。size_t表示加入时的索引。
测试代码:
dispatch_apply(10, globalQueue, ^(size_t index) {
LOG(@"dispatch_apply : %zu", index);
});
2016-09-30 16:57:33.023 base[33146:8699303] dispatch_apply : 0 2016-09-30 16:57:33.023 base[33146:8699545] dispatch_apply : 1 2016-09-30 16:57:33.023 base[33146:8699530] dispatch_apply : 2 2016-09-30 16:57:33.024 base[33146:8699631] dispatch_apply : 3 2016-09-30 16:57:33.024 base[33146:8699303] dispatch_apply : 4 2016-09-30 16:57:33.024 base[33146:8699303] dispatch_apply : 8 2016-09-30 16:57:33.024 base[33146:8699303] dispatch_apply : 9 2016-09-30 16:57:33.024 base[33146:8699545] dispatch_apply : 5 2016-09-30 16:57:33.024 base[33146:8699530] dispatch_apply : 6 2016-09-30 16:57:33.024 base[33146:8699631] dispatch_apply : 7
void dispatch_once(dispatch_once_t *predicate, dispatch_block_t block);
只执行block中代码一次。特别适用于单例模式。
使用示例:
static dispatch_once_t onceToken;
dispatch_once(&onceToken, ^{
});
void dispatch_group_async(dispatch_group_t group, dispatch_queue_t queue, dispatch_block_t block);
该方法将加入队列的block分组管理,特别适用于多方法并行,最后整合全部结果。第一个参数由如下方法得到:
dispatch_group_t dispatch_group_create(void);
void dispatch_group_notify(dispatch_group_t group, dispatch_queue_t queue, dispatch_block_t block);
该方法用于group中所有任务都完成后,回调block内容。
示例代码:
dispatch_group_t group = dispatch_group_create(); dispatch_group_async(group, globalQueue, ^{ LOG(@"dispatch_group_async : 1"); }); dispatch_group_async(group, globalQueue, ^{ LOG(@"dispatch_group_async : 2"); }); dispatch_group_async(group, globalQueue, ^{ LOG(@"dispatch_group_async : 3"); }); dispatch_group_async(group, globalQueue, ^{ LOG(@"dispatch_group_async : 4"); }); dispatch_group_notify(group, globalQueue, ^{ LOG(@"dispatch_group_async : completion"); }); 2016-09-30 17:19:15.490 base[33322:8718096] dispatch_group_async : 1 2016-09-30 17:19:15.490 base[33322:8718097] dispatch_group_async : 3 2016-09-30 17:19:15.490 base[33322:8718042] dispatch_group_async : 2 2016-09-30 17:19:15.490 base[33322:8718098] dispatch_group_async : 4 2016-09-30 17:19:15.491 base[33322:8718098] dispatch_group_async : completion
以上记录了GCD的常见使用方法和示例代码。