GCD底层分析（一）：队列、同步异步函数

2021-08-12 本文已影响0人 HotPotCat

一、GCD 简介

1.1 GCD

GCD（Grand Central Dispatch）本质是 将任务添加到队列，并且指定执行任务的函数。

GCD是纯C语言实现，提供了非常强大的函数。GCD的优势：

是苹果公司为多核的并行运算提出的解决方案。
会自动利用更多的CPU内核(比如双核、四核)。
会自动管理线程的生命周期(创建线程、调度任务、销毁线程)。
程序员只需要告诉 GCD 想要执行什么任务，不需要编写任何线程管理代码。

最简单的一个例子：

dispatch_block_t block = ^{
    NSLog(@"Hello GCD");
};
//串行队列
dispatch_queue_t quene = dispatch_queue_create("com.HotpotCat.zai", NULL);
//函数
dispatch_async(quene, block);

任务使用block封装，这个block块就是任务。任务的block没有参数也没有返回值
quene是创建的串行队列。
通过函数将任务和队列关联在一起。

1.2 GCD 的作用

GCD功能

二、函数与队列

2.1 同步与异步

dispatch_async异步执行任务的函数。
- 不用等待当前语句执行完毕，就可以执行下一条语句。
- 会开启线程执行block的任务。
- 异步是多线程的代名词。
dispatch_sync同步函数。
- 必须等待当前语句执行完毕，才会执行下一条语句。
- 不会开启线程。
- 在当前执行block任务。
block块是在函数内部执行的。

有如下案例：

- (void)test {
    CFAbsoluteTime time = CFAbsoluteTimeGetCurrent();
    dispatch_queue_t queue = dispatch_queue_create("com.HotpotCat.zai", DISPATCH_QUEUE_SERIAL);

    dispatch_async(queue, ^{
        NSLog(@"1: %f",CFAbsoluteTimeGetCurrent() - time);
        method();
    });

    dispatch_sync(queue, ^{
        NSLog(@"2: %f",CFAbsoluteTimeGetCurrent() - time);
        method();
    });

    method();
    NSLog(@"3: %f",CFAbsoluteTimeGetCurrent() - time);
}

void method() {
    sleep(3);
}

输出：

1: 0.000055
2: 3.000264
3: 9.001459

说明同步异步是一个耗时的操作。

2.2 串行队列与并发队列

image.png

队列：是一种数据结构，支持FIFO原则。
串行队列：一次只能进一个任务，任务之间需要排队，DFQ_WIDTH = 1。在上图中任务一比任务二先执行，队列中的任务按顺序执行。
并发队列：一次能调度多个任务（调度多个并不是执行多个，队列不具备执行任务能力，线程才能执行任务），任务一先调度不一定比任务二先执行，得看线程池的调度情况（先调度不一定先执行）。

⚠️ 队列与线程没有任何关系，队列存储任务，线程执行任务。

2.2.1 案例1

有如下代码：

dispatch_queue_t queue = dispatch_queue_create("com.HotpotCat.zai", DISPATCH_QUEUE_CONCURRENT);
NSLog(@"1");
dispatch_async(queue, ^{
    NSLog(@"2");
    dispatch_sync(queue, ^{
        NSLog(@"3");
    });
    NSLog(@"4");
});
NSLog(@"5");

输出：1 5 2 3 4。
分析：queue是并发队列，dispatch_async是异步函数。所以先执行1 5，在dispatch_async的block内部先执行2，由于dispatch_sync同步函数导致3执行完才能执行4。所以输出1 5 2 3 4。

2.2.2 案例2

将上面的例子中DISPATCH_QUEUE_CONCURRENT改为NULL(DISPATCH_QUEUE_SERIAL)：

image.png
这个时候在执行到dispatch_sync的时候发生了死锁。
对于同步函数会进行护犊子，堵塞的是block之后的代码。queue中的任务如下（这里为了简单没有写外层异步函数的任务块）：

image.png
由于queue是串行队列并且支持FIFO，在queue中块任务为同步函数需要保证任务3执行，但是queue是串行队列，任务3的执行依赖于任务4，而任务4以来块任务所以发生了循环等待，造成死锁。如果改为并发队列（3和4可以一起执行）或者任务3为异步函数调用则就不会发生死锁了。

2.2.3 案例3

继续修改代码，将任务4删除，代码如下：

- (void)test{
    dispatch_queue_t queue = dispatch_queue_create("com.HotpotCat.zai", DISPATCH_QUEUE_SERIAL);
    NSLog(@"1");
    dispatch_async(queue, ^{
        NSLog(@"2");
        dispatch_sync(queue, ^{
            NSLog(@"3");
        });
    });
    NSLog(@"5");
}

这个时候仍然发生了死锁。对于queue而言有两个任务块以及一个任务2。

image.png

任务块2阻塞了任务块1的执行完毕，任务块2的执行依赖于任务3的执行，任务3的执行完毕以来于任务块1。这样就造成了死锁。

2.2.4 案例4

- (void)test {
    dispatch_queue_t queue = dispatch_queue_create("com.HotpotCat.zai", DISPATCH_QUEUE_CONCURRENT);

    dispatch_async(queue, ^{
        NSLog(@"1");
    });
    
    dispatch_async(queue, ^{
        NSLog(@"2");
    });

    dispatch_sync(queue, ^{ NSLog(@"3"); });
    
    NSLog(@"0");

    dispatch_async(queue, ^{
        NSLog(@"7");
    });
    dispatch_async(queue, ^{
        NSLog(@"8");
    });
    dispatch_async(queue, ^{
        NSLog(@"9");
    });

}

输出选项：

queue是并发队列，任务3是同步函数执行的。所以任务3后面的任务会被阻塞。那么任务0肯定在任务3之后执行，任务7、8、9肯定在任务0之后执行。所以就有1、 2、 3 — 0 — 7、 8、 9。而由于本身是并发队列，所以1、2、3之间是无序的，7、8、9之间也是无序的。所以A、C符合。

dispatch_async中增加耗时操作，修改如下：

- (void)test {
    dispatch_queue_t queue = dispatch_queue_create("com.HotpotCat.zai", DISPATCH_QUEUE_CONCURRENT);

    dispatch_async(queue, ^{
        sleep(8);
        NSLog(@"1");
    });
    
    dispatch_async(queue, ^{
        sleep(7);
        NSLog(@"2");
    });

    dispatch_sync(queue, ^{
        sleep(1);
        NSLog(@"3");
    });
    
    NSLog(@"0");

    dispatch_async(queue, ^{
        sleep(3);
        NSLog(@"7");
    });
    dispatch_async(queue, ^{
        sleep(2);
        NSLog(@"8");
    });
    dispatch_async(queue, ^{
        sleep(1);
        NSLog(@"9");
    });
}

当任务中有耗时操作时1、2就不一定在0之前执行了，核心点在于dispatch_sync只能保证任务3先执行不能保证1、2先执行。而0仍然在7、8、9之前。
所以只能保证：3在0之前执行。0在7、8、9之前执行（1、2、7、8、9无序，3与1、2之间无序）。此时上面的代码输出3098721。

修改queue为串行队列DISPATCH_QUEUE_SERIAL后，1、2、3以及7、8、9就是按顺序执行了1230789就选A了。

小结：

同步函数串行队列：
- 不会开启线程，在当前线程执行任务。
- 任务串行执行，一个接着一个。
- 会产生阻塞。
同步函数并发队列：
- 不会开启线程，在当前线程执行任务。
- 任务一个接着一个。
异步函数串行队列：
- 开启线程，一条新线程。
- 任务一个接着一个。
异步函数并发队列：
- 开启线程，在当前线程执行任务。
- 任务异步执行，没有顺序，与CPU调用有关。

三、主队列与全局并发队列

通过GCD创建队列，一般有以下4种方式：

- (void)test {
    //串行队列
    dispatch_queue_t serial = dispatch_queue_create("com.HotpotCat.cat", DISPATCH_QUEUE_SERIAL);
    //并发队列
    dispatch_queue_t concurrent = dispatch_queue_create("com.HotpotCat.zai", DISPATCH_QUEUE_CONCURRENT);
    //主队列
    dispatch_queue_t mainQueue = dispatch_get_main_queue();
    //全局队列
    dispatch_queue_t globalQueue = dispatch_get_global_queue(0, 0);
    NSLog(@"serial:%@\nconcurrent:%@\nmainQueue:%@\nglobalQueue:%@",serial,concurrent,mainQueue,globalQueue);
}

3.1 主队列

dispatch_get_main_queue声明如下：

image.png

是一个特殊的串行队列。绑定在UI线程，
在main()之前被自动创建。
image.png

3.1.1 dispatch_get_main_queue 源码分析

那么dispatch_get_main_queue具体是在什么时机创建的呢？
在main_queue的block中打断点bt查看堆栈定位到调用是在libdispatch中：

image.png

在opensource中下载最新的libdispatch-1271.120.2版本源码。

libdispatch源码注释比较少，宏定义很多。

dispatch_get_main_queue
dispatch_get_main_queue定义如下：

dispatch_queue_main_t
dispatch_get_main_queue(void)
{
    //dispatch_queue_main_t 是类型，真正的对象是_dispatch_main_q
    return DISPATCH_GLOBAL_OBJECT(dispatch_queue_main_t, _dispatch_main_q);
}

dispatch_get_main_queue返回了DISPATCH_GLOBAL_OBJECT，参数是dispatch_queue_main_t以及_dispatch_main_q。

DISPATCH_GLOBAL_OBJECT的宏定义如下：

#define DISPATCH_GLOBAL_OBJECT(type, object) ((OS_OBJECT_BRIDGE type)&(object))

可以看到dispatch_queue_main_t是一个类型，真正的对象是object也就是_dispatch_main_q

_dispatch_main_q
_dispatch_main_q的函数搜索不到，此时通过赋值可以直接定位到：

image.png

当然也可以通过label com.apple.main-thread搜索定位到。

struct dispatch_queue_static_s _dispatch_main_q = {
    DISPATCH_GLOBAL_OBJECT_HEADER(queue_main),
#if !DISPATCH_USE_RESOLVERS
    .do_targetq = _dispatch_get_default_queue(true),
#endif
    .dq_state = DISPATCH_QUEUE_STATE_INIT_VALUE(1) |
            DISPATCH_QUEUE_ROLE_BASE_ANON,
    .dq_label = "com.apple.main-thread",
    //DQF_WIDTH(1) 串行队列
    .dq_atomic_flags = DQF_THREAD_BOUND | DQF_WIDTH(1),
    .dq_serialnum = 1,
};

DISPATCH_GLOBAL_OBJECT_HEADER(queue_main)传递的参数是queue_main。
DQF_WIDTH区分串行并行队列。并不是通过serialnum。
最终返回的类型是dispatch_queue_main_t。在使用过程中都是使用dispatch_queue_t接收的。

dispatch_queue_static_s定义如下：

typedef struct dispatch_lane_s {
    DISPATCH_LANE_CLASS_HEADER(lane);
    /* 32bit hole on LP64 */
} DISPATCH_ATOMIC64_ALIGN *dispatch_lane_t;

// Cache aligned type for static queues (main queue, manager)
struct dispatch_queue_static_s {
    struct dispatch_lane_s _as_dl[0]; \
    DISPATCH_LANE_CLASS_HEADER(lane);
} DISPATCH_CACHELINE_ALIGN;

内部实际上是dispatch_lane_s。

3.2 全局队列

dispatch_get_global_queue的实现如下：

dispatch_queue_global_t
dispatch_get_global_queue(intptr_t priority, uintptr_t flags)
{
    dispatch_assert(countof(_dispatch_root_queues) ==
            DISPATCH_ROOT_QUEUE_COUNT);

    //过量使用直接返回0
    if (flags & ~(unsigned long)DISPATCH_QUEUE_OVERCOMMIT) {
        return DISPATCH_BAD_INPUT;
    }
    //根据优先级返回qos
    dispatch_qos_t qos = _dispatch_qos_from_queue_priority(priority);
#if !HAVE_PTHREAD_WORKQUEUE_QOS
    if (qos == QOS_CLASS_MAINTENANCE) {
        qos = DISPATCH_QOS_BACKGROUND;
    } else if (qos == QOS_CLASS_USER_INTERACTIVE) {
        qos = DISPATCH_QOS_USER_INITIATED;
    }
#endif
    if (qos == DISPATCH_QOS_UNSPECIFIED) {
        return DISPATCH_BAD_INPUT;
    }
    //调用 _dispatch_get_root_queue 
    return _dispatch_get_root_queue(qos, flags & DISPATCH_QUEUE_OVERCOMMIT);
}

返回dispatch_queue_global_t。
通过_dispatch_get_root_queue获取队列。

_dispatch_get_root_queue实现：

static inline dispatch_queue_global_t
_dispatch_get_root_queue(dispatch_qos_t qos, bool overcommit)
{
    if (unlikely(qos < DISPATCH_QOS_MIN || qos > DISPATCH_QOS_MAX)) {
        DISPATCH_CLIENT_CRASH(qos, "Corrupted priority");
    }
    return &_dispatch_root_queues[2 * (qos - 1) + overcommit];
}

先对优先级进行验证。
去_dispatch_root_queues集合中取数据。

_dispatch_root_queues实现：

//静态变量集合，随时调用随时取。
struct dispatch_queue_global_s _dispatch_root_queues[] = {
#define _DISPATCH_ROOT_QUEUE_IDX(n, flags) \
        ((flags & DISPATCH_PRIORITY_FLAG_OVERCOMMIT) ? \
        DISPATCH_ROOT_QUEUE_IDX_##n##_QOS_OVERCOMMIT : \
        DISPATCH_ROOT_QUEUE_IDX_##n##_QOS)
#define _DISPATCH_ROOT_QUEUE_ENTRY(n, flags, ...) \
    [_DISPATCH_ROOT_QUEUE_IDX(n, flags)] = { \
        DISPATCH_GLOBAL_OBJECT_HEADER(queue_global), \
        .dq_state = DISPATCH_ROOT_QUEUE_STATE_INIT_VALUE, \
        .do_ctxt = _dispatch_root_queue_ctxt(_DISPATCH_ROOT_QUEUE_IDX(n, flags)), \
        .dq_atomic_flags = DQF_WIDTH(DISPATCH_QUEUE_WIDTH_POOL), \
        .dq_priority = flags | ((flags & DISPATCH_PRIORITY_FLAG_FALLBACK) ? \
                _dispatch_priority_make_fallback(DISPATCH_QOS_##n) : \
                _dispatch_priority_make(DISPATCH_QOS_##n, 0)), \
        __VA_ARGS__ \
    }
    _DISPATCH_ROOT_QUEUE_ENTRY(MAINTENANCE, 0,
        .dq_label = "com.apple.root.maintenance-qos",
        .dq_serialnum = 4,
    ),
    _DISPATCH_ROOT_QUEUE_ENTRY(MAINTENANCE, DISPATCH_PRIORITY_FLAG_OVERCOMMIT,
        .dq_label = "com.apple.root.maintenance-qos.overcommit",
        .dq_serialnum = 5,
    ),
    _DISPATCH_ROOT_QUEUE_ENTRY(BACKGROUND, 0,
        .dq_label = "com.apple.root.background-qos",
        .dq_serialnum = 6,
    ),
    _DISPATCH_ROOT_QUEUE_ENTRY(BACKGROUND, DISPATCH_PRIORITY_FLAG_OVERCOMMIT,
        .dq_label = "com.apple.root.background-qos.overcommit",
        .dq_serialnum = 7,
    ),
    _DISPATCH_ROOT_QUEUE_ENTRY(UTILITY, 0,
        .dq_label = "com.apple.root.utility-qos",
        .dq_serialnum = 8,
    ),
    _DISPATCH_ROOT_QUEUE_ENTRY(UTILITY, DISPATCH_PRIORITY_FLAG_OVERCOMMIT,
        .dq_label = "com.apple.root.utility-qos.overcommit",
        .dq_serialnum = 9,
    ),
    _DISPATCH_ROOT_QUEUE_ENTRY(DEFAULT, DISPATCH_PRIORITY_FLAG_FALLBACK,
        .dq_label = "com.apple.root.default-qos",
        .dq_serialnum = 10,
    ),
    _DISPATCH_ROOT_QUEUE_ENTRY(DEFAULT,
            DISPATCH_PRIORITY_FLAG_FALLBACK | DISPATCH_PRIORITY_FLAG_OVERCOMMIT,
        .dq_label = "com.apple.root.default-qos.overcommit",
        .dq_serialnum = 11,
    ),
    _DISPATCH_ROOT_QUEUE_ENTRY(USER_INITIATED, 0,
        .dq_label = "com.apple.root.user-initiated-qos",
        .dq_serialnum = 12,
    ),
    _DISPATCH_ROOT_QUEUE_ENTRY(USER_INITIATED, DISPATCH_PRIORITY_FLAG_OVERCOMMIT,
        .dq_label = "com.apple.root.user-initiated-qos.overcommit",
        .dq_serialnum = 13,
    ),
    _DISPATCH_ROOT_QUEUE_ENTRY(USER_INTERACTIVE, 0,
        .dq_label = "com.apple.root.user-interactive-qos",
        .dq_serialnum = 14,
    ),
    _DISPATCH_ROOT_QUEUE_ENTRY(USER_INTERACTIVE, DISPATCH_PRIORITY_FLAG_OVERCOMMIT,
        .dq_label = "com.apple.root.user-interactive-qos.overcommit",
        .dq_serialnum = 15,
    ),
};

根据下标取对应的队列。
_dispatch_root_queues是一个静态变量集合，随时调用随时获取。
在使用过程中都是使用dispatch_queue_t接收。
DISPATCH_GLOBAL_OBJECT_HEADER(queue_global)传递的参数是queue_global。

小结：
主队列（dispatch_get_main_queue()）

专⻔用来在主线程上调度任务的串行队列。
不会开启线程。
如果当前主线程正在有任务执行，那么无论主队列中当前被添加了什么任务，都不会被调度。

全局并发队列

为了方便使用，提供了全局队列 dispatch_get_global_queue(0, 0)。
全局队列是一个并发队列。
在使用多线程开发时，如果对队列没有特殊需求，在执行异步任务时，可以直接使用全局队列。

四、dispatch_queue_create

普通的串行队列以及并发队列是通过dispatch_queue_create创建的，它的实现如下：

dispatch_queue_t
dispatch_queue_create(const char *label, dispatch_queue_attr_t attr)
{
    return _dispatch_lane_create_with_target(label, attr,
            DISPATCH_TARGET_QUEUE_DEFAULT, true);
}

返回dispatch_queue_t类型，内部直接调用_dispatch_lane_create_with_target。

4.1 _dispatch_lane_create_with_target

static dispatch_queue_t
_dispatch_lane_create_with_target(const char *label, dispatch_queue_attr_t dqa,
        dispatch_queue_t tq, bool legacy)
{
    //优先级配置，串行直接返回，并行设置。
    dispatch_queue_attr_info_t dqai = _dispatch_queue_attr_to_info(dqa);

    // Step 1: Normalize arguments (qos, overcommit, tq)  优先级相关处理

    ……

    // Step 2: Initialize the queue 初始化队列

    ……

    const void *vtable;
    dispatch_queue_flags_t dqf = legacy ? DQF_MUTABLE : 0;
    if (dqai.dqai_concurrent) {
        //并行，获取class
        vtable = DISPATCH_VTABLE(queue_concurrent);
    } else {
        //串行
        vtable = DISPATCH_VTABLE(queue_serial);
    }

    ……

    //创建dq，申请开辟内存。用 dispatch_lane_t 接收，并不是queue
    dispatch_lane_t dq = _dispatch_object_alloc(vtable,
            sizeof(struct dispatch_lane_s));
    //初始化，dqai_concurrent 表示是否并发，并发传递DISPATCH_QUEUE_WIDTH_MAX ，串行传递1
    _dispatch_queue_init(dq, dqf, dqai.dqai_concurrent ?
            DISPATCH_QUEUE_WIDTH_MAX : 1, DISPATCH_QUEUE_ROLE_INNER |
            (dqai.dqai_inactive ? DISPATCH_QUEUE_INACTIVE : 0));
    //label赋值
    dq->dq_label = label;
    //优先级设置
    dq->dq_priority = _dispatch_priority_make((dispatch_qos_t)dqai.dqai_qos,
            dqai.dqai_relpri);
    if (overcommit == _dispatch_queue_attr_overcommit_enabled) {
        dq->dq_priority |= DISPATCH_PRIORITY_FLAG_OVERCOMMIT;
    }
    if (!dqai.dqai_inactive) {
        _dispatch_queue_priority_inherit_from_target(dq, tq);
        _dispatch_lane_inherit_wlh_from_target(dq, tq);
    }
    _dispatch_retain(tq);
    dq->do_targetq = tq;
    _dispatch_object_debug(dq, "%s", __func__);
    //返回_dispatch_trace_queue_create 的调用的_dq。trace 追踪
    return _dispatch_trace_queue_create(dq)._dq;
}

_dispatch_queue_attr_to_info根据dqa（串行/并行）创建dqai。
返回的dqai进行优先级相关的处理，进行准备工作。
初始化队列
- 根据dqai_concurrent串行并行获取vtable(class name)。参数是queue_concurrent或者queue_serial。
- _dispatch_object_alloc开辟空间。
- _dispatch_queue_init初始化，根据dqai_concurrent传递DISPATCH_QUEUE_WIDTH_MAX 或者 1区分是串行还是并行。
- label赋值。
- 优先级设置。
_dispatch_trace_queue_create追踪返回_dq，重点在dq。

这里创建queue并不是通过queue相关对象进行创建和接收。

4.2 _dispatch_queue_attr_to_info

dispatch_queue_attr_info_t
_dispatch_queue_attr_to_info(dispatch_queue_attr_t dqa)
{
    dispatch_queue_attr_info_t dqai = { };
    //串行队列(dqa 为空)直接返回。
    if (!dqa) return dqai;
……
    //并行队列配置
    size_t idx = (size_t)(dqa - _dispatch_queue_attrs);
    dqai.dqai_inactive = (idx % DISPATCH_QUEUE_ATTR_INACTIVE_COUNT);
    idx /= DISPATCH_QUEUE_ATTR_INACTIVE_COUNT;

    dqai.dqai_concurrent = !(idx % DISPATCH_QUEUE_ATTR_CONCURRENCY_COUNT);
    idx /= DISPATCH_QUEUE_ATTR_CONCURRENCY_COUNT;

    dqai.dqai_relpri = -(int)(idx % DISPATCH_QUEUE_ATTR_PRIO_COUNT);
    idx /= DISPATCH_QUEUE_ATTR_PRIO_COUNT;

    dqai.dqai_qos = idx % DISPATCH_QUEUE_ATTR_QOS_COUNT;
    idx /= DISPATCH_QUEUE_ATTR_QOS_COUNT;

    dqai.dqai_autorelease_frequency =
            idx % DISPATCH_QUEUE_ATTR_AUTORELEASE_FREQUENCY_COUNT;
    idx /= DISPATCH_QUEUE_ATTR_AUTORELEASE_FREQUENCY_COUNT;

    dqai.dqai_overcommit = idx % DISPATCH_QUEUE_ATTR_OVERCOMMIT_COUNT;
    idx /= DISPATCH_QUEUE_ATTR_OVERCOMMIT_COUNT;

    return dqai;
}

串行队列（dqa 为空）直接返回。
并行队列进行一系列配置。

4.3 _dispatch_object_alloc

_dispatch_object_alloc传递的参数是vtable以及sizeof(struct dispatch_lane_s)，那么意味着实际上也是dispatch_lane_s类型。

void *
_dispatch_object_alloc(const void *vtable, size_t size)
{
#if OS_OBJECT_HAVE_OBJC1
    const struct dispatch_object_vtable_s *_vtable = vtable;
    dispatch_object_t dou;
    dou._os_obj = _os_object_alloc_realized(_vtable->_os_obj_objc_isa, size);
    dou._do->do_vtable = vtable;
    return dou._do;
#else
    return _os_object_alloc_realized(vtable, size);
#endif
}

调用_os_object_alloc_realized申请开辟内存。

4.3.1 _os_object_alloc_realized

inline _os_object_t
_os_object_alloc_realized(const void *cls, size_t size)
{
    _os_object_t obj;
    dispatch_assert(size >= sizeof(struct _os_object_s));
    while (unlikely(!(obj = calloc(1u, size)))) {
        _dispatch_temporary_resource_shortage();
    }
    obj->os_obj_isa = cls;
    return obj;
}

内部直接调用calloc开辟空间。

4.4 _dispatch_queue_init

根据是否并行队列传递DISPATCH_QUEUE_WIDTH_MAX与1。

#define DISPATCH_QUEUE_WIDTH_FULL           0x1000ull
#define DISPATCH_QUEUE_WIDTH_POOL (DISPATCH_QUEUE_WIDTH_FULL - 1)
#define DISPATCH_QUEUE_WIDTH_MAX  (DISPATCH_QUEUE_WIDTH_FULL - 2)

static inline dispatch_queue_class_t
_dispatch_queue_init(dispatch_queue_class_t dqu, dispatch_queue_flags_t dqf,
        uint16_t width, uint64_t initial_state_bits)
{
    uint64_t dq_state = DISPATCH_QUEUE_STATE_INIT_VALUE(width);
    dispatch_queue_t dq = dqu._dq;

……

    dq_state |= initial_state_bits;
    dq->do_next = DISPATCH_OBJECT_LISTLESS;
    //设置DQF_WIDTH，区分串行/并行
    dqf |= DQF_WIDTH(width);
    os_atomic_store2o(dq, dq_atomic_flags, dqf, relaxed);
    dq->dq_state = dq_state;
    //标识
    dq->dq_serialnum =
            os_atomic_inc_orig(&_dispatch_queue_serial_numbers, relaxed);
    return dqu;
}

根据width配置状态。
根据width配置队列串行/并行以及dq_serialnum标识。dq_serialnum通过_dispatch_queue_serial_numbers赋值。

unsigned long volatile _dispatch_queue_serial_numbers =
        DISPATCH_QUEUE_SERIAL_NUMBER_INIT;
// skip zero
// 1 - main_q
// 2 - mgr_q
// 3 - mgr_root_q
// 4,5,6,7,8,9,10,11,12,13,14,15 - global queues
// 17 - workloop_fallback_q
// we use 'xadd' on Intel, so the initial value == next assigned
#define DISPATCH_QUEUE_SERIAL_NUMBER_INIT 17
extern unsigned long volatile _dispatch_queue_serial_numbers;

0跳过。
1主队列。
2管理队列。
3管理队列的目标队列。
4~15全局队列。根据qos优先级指定不同队列。
17自动创建相关返回队列。

os_atomic_inc_orig传递的参数是_dispatch_queue_serial_numbers（p）以及relaxed（m）：

#define os_atomic_inc_orig(p, m) \
        os_atomic_add_orig((p), 1, m)

#define os_atomic_add_orig(p, v, m) \
        _os_atomic_c11_op_orig((p), (v), m, add, +)

//##是连接符号，编译后会被去掉
#define _os_atomic_c11_op_orig(p, v, m, o, op) \
        atomic_fetch_##o##_explicit(_os_atomic_c11_atomic(p), v, \
        memory_order_##m)

最终调用的是atomic_fetch_add_explicit(_os_atomic_c11_atomic(17), 1, memory_order_relaxed)。原子替换（p + 1 -> p)，并返回p之前的值。也就相当于是i++这是原子操作。
这里多层宏定义处理是为了兼容不同的c/c++版本。

atomic_fetch_add_explicit参考原子操作atomic_fetch_add

C atomic_fetch_add_explicit(volatile A * obj，M arg，memory_order order);

enum memory_order {
memory_order_relaxed, //不对执行顺序做保证，只保证此操作是原子的
memory_order_consume, // 本线程中,所有后续的有关本原子类型的操作,必须>在本条原子操作完成之后执行
memory_order_acquire, //本线程中,所有后续的读操作必须在本条原子操作完>成后执行
memory_order_release, //本线程中,所有之前的写操作完成后才能执行本条原子操作
memory_order_acq_rel, //同时包含 memory_order_acquire 和 memory_order_release
memory_order_seq_cst //全部存取都按顺序执行
};

4.5 _dispatch_trace_queue_create

#define _dispatch_trace_queue_create _dispatch_introspection_queue_create

dispatch_queue_class_t
_dispatch_introspection_queue_create(dispatch_queue_t dq)
{
    dispatch_queue_introspection_context_t dqic;
    size_t sz = sizeof(struct dispatch_queue_introspection_context_s);

    if (!_dispatch_introspection.debug_queue_inversions) {
        sz = offsetof(struct dispatch_queue_introspection_context_s,
                __dqic_no_queue_inversion);
    }
    dqic = _dispatch_calloc(1, sz);
    dqic->dqic_queue._dq = dq;
    if (_dispatch_introspection.debug_queue_inversions) {
        LIST_INIT(&dqic->dqic_order_top_head);
        LIST_INIT(&dqic->dqic_order_bottom_head);
    }
    dq->do_introspection_ctxt = dqic;

    _dispatch_unfair_lock_lock(&_dispatch_introspection.queues_lock);
    LIST_INSERT_HEAD(&_dispatch_introspection.queues, dqic, dqic_list);
    _dispatch_unfair_lock_unlock(&_dispatch_introspection.queues_lock);

    DISPATCH_INTROSPECTION_INTERPOSABLE_HOOK_CALLOUT(queue_create, dq);
    if (DISPATCH_INTROSPECTION_HOOK_ENABLED(queue_create)) {
        _dispatch_introspection_queue_create_hook(dq);
    }
    return upcast(dq)._dqu;
}

相关跟踪配置。

队列创建流程

五、dispatch_queue_t

由于所有的队列都是通过dispatch_queue_t来接收的，直接研究dispatch_queue_t是一个不错的入口。
dispatch_queue_t的点击会跳转到：

DISPATCH_DECL(dispatch_queue);

5.1 DISPATCH_DECL 源码分析

DISPATCH_DECL宏定义：

#define DISPATCH_DECL(name) OS_OBJECT_DECL_SUBCLASS(name, dispatch_object)

真实类型是OS_OBJECT_DECL_SUBCLASS，在源码中它的定义有两个，一个是针对objc的，定义如下：

#define OS_OBJECT_DECL_SUBCLASS(name, super) \
        OS_OBJECT_DECL_IMPL(name, NSObject, <OS_OBJECT_CLASS(super)>)

OS_OBJECT_DECL_IMPL定义如下：

#define OS_OBJECT_DECL_IMPL(name, adhere, ...) \
        OS_OBJECT_DECL_PROTOCOL(name, __VA_ARGS__) \
        typedef adhere<OS_OBJECT_CLASS(name)> \
                * OS_OBJC_INDEPENDENT_CLASS name##_t

OS_OBJECT_DECL_PROTOCOL定义如下：

#define OS_OBJECT_DECL_PROTOCOL(name, ...) \
        @protocol OS_OBJECT_CLASS(name) __VA_ARGS__ \
        @end

也就是定义了一个@protocol。

OS_OBJC_INDEPENDENT_CLASS定义如下：

#if __has_attribute(objc_independent_class)
#define OS_OBJC_INDEPENDENT_CLASS __attribute__((objc_independent_class))
#endif // __has_attribute(objc_independent_class)
#ifndef OS_OBJC_INDEPENDENT_CLASS
#define OS_OBJC_INDEPENDENT_CLASS
#endif

为了方便分析，这里假设它走的是OS_OBJC_INDEPENDENT_CLASS为空。

OS_OBJECT_CLASS定义如下：

#define OS_OBJECT_CLASS(name) OS_##name

本质上是OS_拼接。

在整个宏定义中参数 name: dispatch_queue super: dispatch_object，整个宏定义替换完成后如下：

@protocol OS_dispatch_queue <OS_dispatch_object>
@end
typedef NSObject<OS_dispatch_queue>  *dispatch_queue_t

当然在源码中搜索#define DISPATCH_DECL可以搜到多个，其中有一个定义如下：

#define DISPATCH_DECL(name) \
        typedef struct name##_s : public dispatch_object_s {} *name##_t

替换后如下：

typedef struct dispatch_queue_s : public dispatch_object_s {} *dispatch_queue_t

dispatch_queue_t是一个结构体来自于dispatch_queue_s。
dispatch_queue_s继承自dispatch_object_s。
dispatch_queue_t -> dispatch_queue_s -> dispatch_object_s类似于class -> objc_class -> objc_object。
本质上dispatch_queue_t是dispatch_queue_s结构体类型。

5.2 dispatch_queue_s 分析

要研究dispatch_queue_t那么就要研究它的类型dispatch_queue_s。
dispatch_queue_s定义如下：

struct dispatch_queue_s {
    DISPATCH_QUEUE_CLASS_HEADER(queue, void *__dq_opaque1);
    /* 32bit hole on LP64 */
} DISPATCH_ATOMIC64_ALIGN;

DISPATCH_QUEUE_CLASS_HEADER宏定义如下：

#define DISPATCH_QUEUE_CLASS_HEADER(x, __pointer_sized_field__) \
    _DISPATCH_QUEUE_CLASS_HEADER(x, __pointer_sized_field__); \
    /* LP64 global queue cacheline boundary */ \
    unsigned long dq_serialnum; \
    const char *dq_label; \
    DISPATCH_UNION_LE(uint32_t volatile dq_atomic_flags, \
        const uint16_t dq_width, \
        const uint16_t __dq_opaque2 \
    ); \
    dispatch_priority_t dq_priority; \
    union { \
        struct dispatch_queue_specific_head_s *dq_specific_head; \
        struct dispatch_source_refs_s *ds_refs; \
        struct dispatch_timer_source_refs_s *ds_timer_refs; \
        struct dispatch_mach_recv_refs_s *dm_recv_refs; \
        struct dispatch_channel_callbacks_s const *dch_callbacks; \
    }; \
    int volatile dq_sref_cnt

内部继承于_DISPATCH_QUEUE_CLASS_HEADER：

#define _DISPATCH_QUEUE_CLASS_HEADER(x, __pointer_sized_field__) \
    DISPATCH_OBJECT_HEADER(x); \
    __pointer_sized_field__; \
    DISPATCH_UNION_LE(uint64_t volatile dq_state, \
            dispatch_lock dq_state_lock, \
            uint32_t dq_state_bits \
    )
#endif

_DISPATCH_QUEUE_CLASS_HEADER内部又继承自DISPATCH_OBJECT_HEADER：

#define DISPATCH_OBJECT_HEADER(x) \
    struct dispatch_object_s _as_do[0]; \
    _DISPATCH_OBJECT_HEADER(x)

这里就对接到了dispatch_object_s类型。

DISPATCH_OBJECT_HEADER内部又使用_DISPATCH_OBJECT_HEADER：

#define _DISPATCH_OBJECT_HEADER(x) \
    struct _os_object_s _as_os_obj[0]; \
    OS_OBJECT_STRUCT_HEADER(dispatch_##x); \
    struct dispatch_##x##_s *volatile do_next; \
    struct dispatch_queue_s *do_targetq; \
    void *do_ctxt; \
    union { \
        dispatch_function_t DISPATCH_FUNCTION_POINTER do_finalizer; \
        void *do_introspection_ctxt; \
    }

也就是最终使用的是_os_object_s类型。

OS_OBJECT_STRUCT_HEADER类型：

#define OS_OBJECT_STRUCT_HEADER(x) \
    _OS_OBJECT_HEADER(\
    const struct x##_vtable_s *__ptrauth_objc_isa_pointer do_vtable, \
    do_ref_cnt, \
    do_xref_cnt)
#endif

_OS_OBJECT_HEADER有3个成员变量：

#define _OS_OBJECT_HEADER(isa, ref_cnt, xref_cnt) \
        isa; /* must be pointer-sized and use __ptrauth_objc_isa_pointer */ \
        int volatile ref_cnt; \
        int volatile xref_cnt

至此整个继承链为：dispatch_queue_t -> dispatch_queue_s -> dispatch_object_s -> _os_object_s。

5.3 dispatch_object_s 分析

dispatch_object_s源码如下：

typedef struct dispatch_object_s {
private:
    dispatch_object_s();
    ~dispatch_object_s();
    dispatch_object_s(const dispatch_object_s &);
    void operator=(const dispatch_object_s &);
} *dispatch_object_t;
#define DISPATCH_DECL(name) \
        typedef struct name##_s : public dispatch_object_s {} *name##_t
#define DISPATCH_DECL_SUBCLASS(name, base) \
        typedef struct name##_s : public base##_s {} *name##_t
#define DISPATCH_GLOBAL_OBJECT(type, object) (static_cast<type>(&(object)))
#define DISPATCH_RETURNS_RETAINED
#else /* Plain C */
#ifndef __DISPATCH_BUILDING_DISPATCH__
typedef union {
    struct _os_object_s *_os_obj;
    struct dispatch_object_s *_do;
    struct dispatch_queue_s *_dq;
    struct dispatch_queue_attr_s *_dqa;
    struct dispatch_group_s *_dg;
    struct dispatch_source_s *_ds;
    struct dispatch_channel_s *_dch;
    struct dispatch_mach_s *_dm;
    struct dispatch_mach_msg_s *_dmsg;
    struct dispatch_semaphore_s *_dsema;
    struct dispatch_data_s *_ddata;
    struct dispatch_io_s *_dchannel;
} dispatch_object_t DISPATCH_TRANSPARENT_UNION;

dispatch_object_t是一个联合体，那么意味着它可以是里面数据类型的任意一种。
其中有dispatch_object_s类型，那么意味者它底层实际上是dispatch_object_t类型。
_os_object_s也与上面的分析相对应。

总结： 整个继承链为：dispatch_queue_t -> dispatch_queue_s -> dispatch_object_s -> _os_object_s -> dispatch_object_t。

队列类型继承图

六、GCD 函数调用流程

对于GCD的block块不是我们主动调用的，那么它是在什么时机调用的呢？任务的执行流程是怎么样的呢？

6.1 同步流程

一般同步任务实现如下：

dispatch_sync(dispatch_get_global_queue(0, 0), ^{
    NSLog(@"test");
});

6.1.1 dispatch_sync 源码分析

void
dispatch_sync(dispatch_queue_t dq, dispatch_block_t work)
{
    uintptr_t dc_flags = DC_FLAG_BLOCK;
    if (unlikely(_dispatch_block_has_private_data(work))) {
        return _dispatch_sync_block_with_privdata(dq, work, dc_flags);
    }
    _dispatch_sync_f(dq, work, _dispatch_Block_invoke(work), dc_flags);
}

#define _dispatch_Block_invoke(bb) \
        ((dispatch_function_t)((struct Block_layout *)bb)->invoke)

核心是_dispatch_sync_f的调用。参数是队列、block、对block包装成invoke、DC_FLAG_BLOCK。
重点就是这个invoke在什么时机被调用。

_dispatch_sync_f：

static void
_dispatch_sync_f(dispatch_queue_t dq, void *ctxt, dispatch_function_t func,
        uintptr_t dc_flags)
{
    _dispatch_sync_f_inline(dq, ctxt, func, dc_flags);
}

_dispatch_sync_f内部只是对_dispatch_sync_f_inline的调用。

6.1.2 _dispatch_sync_f_inline

由于_dispatch_sync_f_inline中有很多分支调用，直接下符号断点跟踪流程：

static inline void
_dispatch_sync_f_inline(dispatch_queue_t dq, void *ctxt,
        dispatch_function_t func, uintptr_t dc_flags)
{
    //串行队列
    if (likely(dq->dq_width == 1)) {
        return _dispatch_barrier_sync_f(dq, ctxt, func, dc_flags);
    }

    if (unlikely(dx_metatype(dq) != _DISPATCH_LANE_TYPE)) {
        DISPATCH_CLIENT_CRASH(0, "Queue type doesn't support dispatch_sync");
    }

    dispatch_lane_t dl = upcast(dq)._dl;
    // Global concurrent queues and queues bound to non-dispatch threads
    // always fall into the slow case, see DISPATCH_ROOT_QUEUE_STATE_INIT_VALUE
    //全局并发队列
    if (unlikely(!_dispatch_queue_try_reserve_sync_width(dl))) {
        return _dispatch_sync_f_slow(dl, ctxt, func, 0, dl, dc_flags);
    }

    if (unlikely(dq->do_targetq->do_targetq)) {
        return _dispatch_sync_recurse(dl, ctxt, func, dc_flags);
    }
    _dispatch_introspection_sync_begin(dl);
    //自定义并发队
    _dispatch_sync_invoke_and_complete(dl, ctxt, func DISPATCH_TRACE_ARG(
            _dispatch_trace_item_sync_push_pop(dq, ctxt, func, dc_flags)));
}

串行队列走_dispatch_barrier_sync_f流程。
全局并发队列走_dispatch_sync_f_slow流程。
自定义并发队列走_dispatch_sync_invoke_and_complete流程。

6.1.3 _dispatch_barrier_sync_f（串行）

一般调用如下：

dispatch_queue_t queue = dispatch_queue_create("test", NULL);
dispatch_sync(queue, ^{
    NSLog(@"test");
});

_dispatch_barrier_sync_f源码实现：

static void
_dispatch_barrier_sync_f(dispatch_queue_t dq, void *ctxt,
        dispatch_function_t func, uintptr_t dc_flags)
{
    _dispatch_barrier_sync_f_inline(dq, ctxt, func, dc_flags);
}

内部是对_dispatch_barrier_sync_f_inline的调用：

static inline void
_dispatch_barrier_sync_f_inline(dispatch_queue_t dq, void *ctxt,
        dispatch_function_t func, uintptr_t dc_flags)
{
    dispatch_tid tid = _dispatch_tid_self();

    if (unlikely(dx_metatype(dq) != _DISPATCH_LANE_TYPE)) {
        DISPATCH_CLIENT_CRASH(0, "Queue type doesn't support dispatch_sync");
    }

    dispatch_lane_t dl = upcast(dq)._dl;

    if (unlikely(!_dispatch_queue_try_acquire_barrier_sync(dl, tid))) {
        return _dispatch_sync_f_slow(dl, ctxt, func, DC_FLAG_BARRIER, dl,
                DC_FLAG_BARRIER | dc_flags);
    }

    if (unlikely(dl->do_targetq->do_targetq)) {
        return _dispatch_sync_recurse(dl, ctxt, func,
                DC_FLAG_BARRIER | dc_flags);
    }
    _dispatch_introspection_sync_begin(dl);
    _dispatch_lane_barrier_sync_invoke_and_complete(dl, ctxt, func
            DISPATCH_TRACE_ARG(_dispatch_trace_item_sync_push_pop(
                    dq, ctxt, func, dc_flags | DC_FLAG_BARRIER)));
}

会调用到_dispatch_lane_barrier_sync_invoke_and_complete：

static void
_dispatch_lane_barrier_sync_invoke_and_complete(dispatch_lane_t dq,
        void *ctxt, dispatch_function_t func DISPATCH_TRACE_ARG(void *dc))
{
    _dispatch_sync_function_invoke_inline(dq, ctxt, func);
    ……
}

其中调用了_dispatch_sync_function_invoke_inline:

static inline void
_dispatch_sync_function_invoke_inline(dispatch_queue_class_t dq, void *ctxt,
        dispatch_function_t func)
{
    dispatch_thread_frame_s dtf;
    _dispatch_thread_frame_push(&dtf, dq);
    _dispatch_client_callout(ctxt, func);
    _dispatch_perfmon_workitem_inc();
    _dispatch_thread_frame_pop(&dtf);
}

调用了_dispatch_client_callout。

在block实现中打断点堆栈如下：

  * frame #0: 0x000000010f94d477 GCDDemo`__29-[ViewController viewDidLoad]_block_invoke(.block_descriptor=0x000000010f950038) at ViewController.m:30:9
    frame #1: 0x000000010fbc39c8 libdispatch.dylib`_dispatch_client_callout + 8
    frame #2: 0x000000010fbd2bfe libdispatch.dylib`_dispatch_lane_barrier_sync_invoke_and_complete + 132
    frame #3: 0x000000010f94d44a GCDDemo`-[ViewController viewDidLoad](self=0x00007f8f32c09ae0, _cmd="viewDidLoad") at ViewController.m:29:5

正好对应上。

6.1.4 _dispatch_sync_f_slow（全局并发）

dispatch_sync(dispatch_get_global_queue(0, 0), ^{
    NSLog(@"test");
});

_dispatch_sync_f_slow源码如下：

static void
_dispatch_sync_f_slow(dispatch_queue_class_t top_dqu, void *ctxt,
        dispatch_function_t func, uintptr_t top_dc_flags,
        dispatch_queue_class_t dqu, uintptr_t dc_flags)
{
    dispatch_queue_t top_dq = top_dqu._dq;
    dispatch_queue_t dq = dqu._dq;
    if (unlikely(!dq->do_targetq)) {
        return _dispatch_sync_function_invoke(dq, ctxt, func);
    }

    pthread_priority_t pp = _dispatch_get_priority();
    struct dispatch_sync_context_s dsc = {
        .dc_flags    = DC_FLAG_SYNC_WAITER | dc_flags,
        .dc_func     = _dispatch_async_and_wait_invoke,
        .dc_ctxt     = &dsc,
        .dc_other    = top_dq,
        .dc_priority = pp | _PTHREAD_PRIORITY_ENFORCE_FLAG,
        .dc_voucher  = _voucher_get(),
        .dsc_func    = func,
        .dsc_ctxt    = ctxt,
        .dsc_waiter  = _dispatch_tid_self(),
    };

    _dispatch_trace_item_push(top_dq, &dsc);
    __DISPATCH_WAIT_FOR_QUEUE__(&dsc, dq);

    if (dsc.dsc_func == NULL) {
        // dsc_func being cleared means that the block ran on another thread ie.
        // case (2) as listed in _dispatch_async_and_wait_f_slow.
        dispatch_queue_t stop_dq = dsc.dc_other;
        return _dispatch_sync_complete_recurse(top_dq, stop_dq, top_dc_flags);
    }

    _dispatch_introspection_sync_begin(top_dq);
    _dispatch_trace_item_pop(top_dq, &dsc);
    _dispatch_sync_invoke_and_complete_recurse(top_dq, ctxt, func,top_dc_flags
            DISPATCH_TRACE_ARG(&dsc));
}

断点继续跟踪走_dispatch_sync_function_invoke逻辑（证明系统全局队列没有do_targetq，自己创建的有。），它是对_dispatch_sync_function_invoke_inline的封装：

static void
_dispatch_sync_function_invoke(dispatch_queue_class_t dq, void *ctxt,
        dispatch_function_t func)
{
    _dispatch_sync_function_invoke_inline(dq, ctxt, func);
}

_dispatch_sync_function_invoke_inline实现如下：

static inline void
_dispatch_sync_function_invoke_inline(dispatch_queue_class_t dq, void *ctxt,
        dispatch_function_t func)
{
    dispatch_thread_frame_s dtf;
    _dispatch_thread_frame_push(&dtf, dq);
    _dispatch_client_callout(ctxt, func);
    _dispatch_perfmon_workitem_inc();
    _dispatch_thread_frame_pop(&dtf);
}

_dispatch_client_callout中传递的参数是block，实现如下：

void
_dispatch_client_callout(void *ctxt, dispatch_function_t f)
{
    _dispatch_get_tsd_base();
    void *u = _dispatch_get_unwind_tsd();
    if (likely(!u)) return f(ctxt);
    _dispatch_set_unwind_tsd(NULL);
    f(ctxt);
    _dispatch_free_unwind_tsd();
    _dispatch_set_unwind_tsd(u);
}

内部直接调用了f(ctxt)，也就是对block的调用。
_dispatch_client_callout的实现有多个，具体实现都是对f(ctxt)的调用。

验证：在block实现内部直接断点，有如下堆栈：

  * frame #0: 0x000000010c232477 GCDDemo`__29-[ViewController viewDidLoad]_block_invoke(.block_descriptor=0x000000010c235038) at ViewController.m:33:9
    frame #1: 0x000000010c4a89c8 libdispatch.dylib`_dispatch_client_callout + 8
    frame #2: 0x000000010c4ad2ab libdispatch.dylib`_dispatch_sync_function_invoke + 127
    frame #3: 0x000000010c232440 GCDDemo`-[ViewController viewDidLoad](self=0x00007fbc29c0b660, _cmd="viewDidLoad") at ViewController.m:32:5

从堆栈验证block最终也是_dispatch_client_callout调用的。

6.1.5 _dispatch_sync_invoke_and_complete（自定义并发）

dispatch_queue_t queue = dispatch_queue_create("test", DISPATCH_QUEUE_CONCURRENT);
dispatch_sync(queue, ^{
    NSLog(@"test");
});

_dispatch_sync_invoke_and_complete源码如下：

static void
_dispatch_sync_invoke_and_complete(dispatch_lane_t dq, void *ctxt,
        dispatch_function_t func DISPATCH_TRACE_ARG(void *dc))
{
    _dispatch_sync_function_invoke_inline(dq, ctxt, func);
    _dispatch_trace_item_complete(dc);
    _dispatch_lane_non_barrier_complete(dq, 0);
}

这里dispatch_function_t func DISPATCH_TRACE_ARG(void *dc)看起来是一个参数（其实有可能是一个参数，有可能是两个参数）对应传过来的func DISPATCH_TRACE_ARG(_dispatch_trace_item_sync_push_pop(dq, ctxt, func, dc_flags))
DISPATCH_TRACE_ARG定义：

#if DISPATCH_USE_DTRACE_INTROSPECTION || DISPATCH_INTROSPECTION
......
#define DISPATCH_TRACE_ARG(arg) , arg
......
#else
......
#define DISPATCH_TRACE_ARG(arg)
......
#endif

也就是dispatch_function_t func,_dispatch_trace_item_sync_push_pop(dq, ctxt, func, dc_flags)或者dispatch_function_t func。后面的参数根据条件是可选参数。

_dispatch_sync_function_invoke_inline：

static inline void
_dispatch_sync_function_invoke_inline(dispatch_queue_class_t dq, void *ctxt,
        dispatch_function_t func)
{
    dispatch_thread_frame_s dtf;
    _dispatch_thread_frame_push(&dtf, dq);
    _dispatch_client_callout(ctxt, func);
    _dispatch_perfmon_workitem_inc();
    _dispatch_thread_frame_pop(&dtf);
}

其中调用了_dispatch_client_callout，block实现打断点堆栈如下：

  * frame #0: 0x000000010a7aa477 GCDDemo`__29-[ViewController viewDidLoad]_block_invoke(.block_descriptor=0x000000010a7ad040) at ViewController.m:29:9
    frame #1: 0x000000010aa209c8 libdispatch.dylib`_dispatch_client_callout + 8
    frame #2: 0x000000010aa302fb libdispatch.dylib`_dispatch_sync_invoke_and_complete + 132
    frame #3: 0x000000010a7aa440 GCDDemo`-[ViewController viewDidLoad](self=0x00007fd5dc506a90, _cmd="viewDidLoad") at ViewController.m:28:5

⚠️异步函数调用并不会开启线程，调用过程中直接执行回调，无论队列串行/并行任务都按序执行。

6.2 异步流程

对于异步函数一般实现：

dispatch_async(dispatch_get_global_queue(0, 0), ^{
    NSLog(@"test");
});

6.2.1 dispatch_async 源码分析

void
dispatch_async(dispatch_queue_t dq, dispatch_block_t work)
{
    dispatch_continuation_t dc = _dispatch_continuation_alloc();
    uintptr_t dc_flags = DC_FLAG_CONSUME;
    dispatch_qos_t qos;

    qos = _dispatch_continuation_init(dc, dq, work, 0, dc_flags);
    _dispatch_continuation_async(dq, dc, qos, dc->dc_flags);
}

内部对work封装成qos。
调用的是_dispatch_continuation_async。

_dispatch_continuation_init的实现：

static inline dispatch_qos_t
_dispatch_continuation_init(dispatch_continuation_t dc,
        dispatch_queue_class_t dqu, dispatch_block_t work,
        dispatch_block_flags_t flags, uintptr_t dc_flags)
{
    void *ctxt = _dispatch_Block_copy(work);

    dc_flags |= DC_FLAG_BLOCK | DC_FLAG_ALLOCATED;
    if (unlikely(_dispatch_block_has_private_data(work))) {
        dc->dc_flags = dc_flags;
        dc->dc_ctxt = ctxt;
        // will initialize all fields but requires dc_flags & dc_ctxt to be set
        return _dispatch_continuation_init_slow(dc, dqu, flags);
    }
    //invoke包装成func
    dispatch_function_t func = _dispatch_Block_invoke(work);
    if (dc_flags & DC_FLAG_CONSUME) {
        func = _dispatch_call_block_and_release;
    }
    return _dispatch_continuation_init_f(dc, dqu, ctxt, func, flags, dc_flags);
}

对work进行拷贝成ctxt以及包装成func。
调用_dispatch_continuation_init_f。

_dispatch_continuation_init_f的实现如下：

static inline dispatch_qos_t
_dispatch_continuation_init_f(dispatch_continuation_t dc,
        dispatch_queue_class_t dqu, void *ctxt, dispatch_function_t f,
        dispatch_block_flags_t flags, uintptr_t dc_flags)
{
    pthread_priority_t pp = 0;
    dc->dc_flags = dc_flags | DC_FLAG_ALLOCATED;
    //将f赋值给dc_func
    dc->dc_func = f;
    //ctxt 赋值给dc_ctxt
    dc->dc_ctxt = ctxt;
    // in this context DISPATCH_BLOCK_HAS_PRIORITY means that the priority
    // should not be propagated, only taken from the handler if it has one
    if (!(flags & DISPATCH_BLOCK_HAS_PRIORITY)) {
        pp = _dispatch_priority_propagate();
    }
    _dispatch_continuation_voucher_set(dc, flags);
    return _dispatch_continuation_priority_set(dc, dqu, pp, flags);
}

将invoke和block赋值给dc对应的字段。
_dispatch_continuation_priority_set进行优先级处理。

_dispatch_continuation_priority_set的实现：

static inline dispatch_qos_t
_dispatch_continuation_priority_set(dispatch_continuation_t dc,
        dispatch_queue_class_t dqu,
        pthread_priority_t pp, dispatch_block_flags_t flags)
{
    dispatch_qos_t qos = DISPATCH_QOS_UNSPECIFIED;
#if HAVE_PTHREAD_WORKQUEUE_QOS
    dispatch_queue_t dq = dqu._dq;

    if (likely(pp)) {
        bool enforce = (flags & DISPATCH_BLOCK_ENFORCE_QOS_CLASS);
        bool is_floor = (dq->dq_priority & DISPATCH_PRIORITY_FLAG_FLOOR);
        bool dq_has_qos = (dq->dq_priority & DISPATCH_PRIORITY_REQUESTED_MASK);
        if (enforce) {
            pp |= _PTHREAD_PRIORITY_ENFORCE_FLAG;
            qos = _dispatch_qos_from_pp_unsafe(pp);
        } else if (!is_floor && dq_has_qos) {
            pp = 0;
        } else {
            qos = _dispatch_qos_from_pp_unsafe(pp);
        }
    }
    dc->dc_priority = pp;
#else
    (void)dc; (void)dqu; (void)pp; (void)flags;
#endif
    return qos;
}

生成qos对优先级进行了配置。

为什么异步函数做了优先级处理？

异步函数代表异步调用，会产生无序调用。优先级是参考依据。
回调必然是异步的。
既然被封装成了qos，那么必然有个时机取出block进行回调。

6.2.2 _dispatch_continuation_async

static inline void
_dispatch_continuation_async(dispatch_queue_class_t dqu,
        dispatch_continuation_t dc, dispatch_qos_t qos, uintptr_t dc_flags)
{
#if DISPATCH_INTROSPECTION
    if (!(dc_flags & DC_FLAG_NO_INTROSPECTION)) {
        _dispatch_trace_item_push(dqu, dc);
    }
#else
    (void)dc_flags;
#endif
    return dx_push(dqu._dq, dc, qos);
}

内部调用了dx_push：

#define dx_push(x, y, z) dx_vtable(x)->dq_push(x, y, z)

qos是第三个参数z，那么核心就是dq_push（dx_vtable判断队列类型），搜索后赋值如下：

image.png
这里就和前面的3.1中queue_main 3.2中queue_global 4.1中 queue_concurrent和queue_serial对应上了：

queue_global：_dispatch_root_queue_push。
queue_main：_dispatch_main_queue_push。
queue_serial：_dispatch_lane_push。
queue_concurrent：_dispatch_lane_concurrent_push。

6.2.2 _dispatch_root_queue_push（全局队列）

void
_dispatch_root_queue_push(dispatch_queue_global_t rq, dispatch_object_t dou,
        dispatch_qos_t qos)
{
#if DISPATCH_USE_KEVENT_WORKQUEUE
    dispatch_deferred_items_t ddi = _dispatch_deferred_items_get();
    if (unlikely(ddi && ddi->ddi_can_stash)) {……}
#endif
#if HAVE_PTHREAD_WORKQUEUE_QOS
    if (_dispatch_root_queue_push_needs_override(rq, qos)) {
        return _dispatch_root_queue_push_override(rq, dou, qos);
    }
#else
    (void)qos;
#endif
    _dispatch_root_queue_push_inline(rq, dou, dou, 1);
}

内部调用了_dispatch_root_queue_push_override：

static void
_dispatch_root_queue_push_override(dispatch_queue_global_t orig_rq,
        dispatch_object_t dou, dispatch_qos_t qos)
{
    bool overcommit = orig_rq->dq_priority & DISPATCH_PRIORITY_FLAG_OVERCOMMIT;
    dispatch_queue_global_t rq = _dispatch_get_root_queue(qos, overcommit);
    dispatch_continuation_t dc = dou._dc;

    if (_dispatch_object_is_redirection(dc)) {
        // no double-wrap is needed, _dispatch_async_redirect_invoke will do
        // the right thing
        dc->dc_func = (void *)orig_rq;
    } else {
        dc = _dispatch_continuation_alloc();
        dc->do_vtable = DC_VTABLE(OVERRIDE_OWNING);
        dc->dc_ctxt = dc;
        dc->dc_other = orig_rq;
        dc->dc_data = dou._do;
        dc->dc_priority = DISPATCH_NO_PRIORITY;
        dc->dc_voucher = DISPATCH_NO_VOUCHER;
    }
    _dispatch_root_queue_push_inline(rq, dc, dc, 1);
}

直接调用_dispatch_root_queue_push_inline：

static inline void
_dispatch_root_queue_push_inline(dispatch_queue_global_t dq,
        dispatch_object_t _head, dispatch_object_t _tail, int n)
{
    struct dispatch_object_s *hd = _head._do, *tl = _tail._do;
    if (unlikely(os_mpsc_push_list(os_mpsc(dq, dq_items), hd, tl, do_next))) {
        //调用_dispatch_root_queue_poke
        return _dispatch_root_queue_poke(dq, n, 0);
    }
}

直接调用_dispatch_root_queue_poke:

void
_dispatch_root_queue_poke(dispatch_queue_global_t dq, int n, int floor)
{
    ……
    return _dispatch_root_queue_poke_slow(dq, n, floor);
}

_dispatch_root_queue_poke_slow实现如下：

static void
_dispatch_root_queue_poke_slow(dispatch_queue_global_t dq, int n, int floor)
{
    ……

    _dispatch_root_queues_init();
    ……
}

_dispatch_root_queues_init实现如下：

static inline void
_dispatch_root_queues_init(void)
{
    dispatch_once_f(&_dispatch_root_queues_pred, NULL,
            _dispatch_root_queues_init_once);
}

是一个单例的调用，调用_dispatch_root_queues_init_once：

image.png

_dispatch_worker_thread2赋值给cfg.workq_cb，交由pthread调用。
交由_pthread_workqueue_init_with_workloop调用，也就是os调用（cpu调度）。

直接在block的实现打断点查看堆栈如下：

* thread #4, queue = 'com.apple.root.default-qos', stop reason = breakpoint 1.1
  * frame #0: 0x0000000104614477 GCDDemo`__29-[ViewController viewDidLoad]_block_invoke(.block_descriptor=0x0000000104617038) at ViewController.m:33:9
    frame #1: 0x00000001048897ec libdispatch.dylib`_dispatch_call_block_and_release + 12
    frame #2: 0x000000010488a9c8 libdispatch.dylib`_dispatch_client_callout + 8
    frame #3: 0x000000010488ce46 libdispatch.dylib`_dispatch_queue_override_invoke + 1032
    frame #4: 0x000000010489c508 libdispatch.dylib`_dispatch_root_queue_drain + 351
    frame #5: 0x000000010489ce6d libdispatch.dylib`_dispatch_worker_thread2 + 135
    frame #6: 0x00007fff611639f7 libsystem_pthread.dylib`_pthread_wqthread + 220
    frame #7: 0x00007fff61162b77 libsystem_pthread.dylib`start_wqthread + 15

可以看到是在_pthread中系统进行调用的。

GCD底层也是封装的pthread。

_dispatch_worker_thread2的实现：

image.png

调用了_dispatch_root_queue_drain：

image.png

是对_dispatch_continuation_pop_inline的调用：

image.png

dx_invoke宏定义就是_dispatch_queue_override_invoke：

image.png

_dispatch_queue_override_invoke内部的调用了_dispatch_continuation_invoke_inline：

image.png

_dispatch_client_callout调用block。

6.2.3 _dispatch_main_queue_push（主队列）

dispatch_async(dispatch_get_main_queue(), ^{
    NSLog(@"test");
});

与全局并发队列同理，当队列为主队列时进入_dispatch_main_queue_push逻辑:

void
_dispatch_main_queue_push(dispatch_queue_main_t dq, dispatch_object_t dou,
        dispatch_qos_t qos)
{
    // Same as _dispatch_lane_push() but without the refcounting due to being
    // a global object
    if (_dispatch_queue_push_item(dq, dou)) {
        return dx_wakeup(dq, qos, DISPATCH_WAKEUP_MAKE_DIRTY);
    }

    qos = _dispatch_queue_push_qos(dq, qos);
    if (_dispatch_queue_need_override(dq, qos)) {
        return dx_wakeup(dq, qos, 0);
    }
}

内部调用了dx_wakeup定义如下：

#define dx_wakeup(x, y, z) dx_vtable(x)->dq_wakeup(x, y, z)

qos是第二个参数，所以核心实现是dq_wakeup：

image.png

实现是_dispatch_main_queue_wakeup：

void
_dispatch_main_queue_wakeup(dispatch_queue_main_t dq, dispatch_qos_t qos,
        dispatch_wakeup_flags_t flags)
{
#if DISPATCH_COCOA_COMPAT
    if (_dispatch_queue_is_thread_bound(dq)) {
        return _dispatch_runloop_queue_wakeup(dq->_as_dl, qos, flags);
    }
#endif
    return _dispatch_lane_wakeup(dq, qos, flags);
}

下符号断点后走的是_dispatch_runloop_queue_wakeup逻辑：

void
_dispatch_runloop_queue_wakeup(dispatch_lane_t dq, dispatch_qos_t qos,
        dispatch_wakeup_flags_t flags)
{
    if (unlikely(_dispatch_queue_atomic_flags(dq) & DQF_RELEASED)) {
        // <rdar://problem/14026816>
        return _dispatch_lane_wakeup(dq, qos, flags);
    }

    if (flags & DISPATCH_WAKEUP_MAKE_DIRTY) {
        os_atomic_or2o(dq, dq_state, DISPATCH_QUEUE_DIRTY, release);
    }
    if (_dispatch_queue_class_probe(dq)) {
        return _dispatch_runloop_queue_poke(dq, qos, flags);
    }

    qos = _dispatch_runloop_queue_reset_max_qos(dq);
    if (qos) {
        mach_port_t owner = DISPATCH_QUEUE_DRAIN_OWNER(dq);
        if (_dispatch_queue_class_probe(dq)) {
            _dispatch_runloop_queue_poke(dq, qos, flags);
        }
        _dispatch_thread_override_end(owner, dq);
        return;
    }
    if (flags & DISPATCH_WAKEUP_CONSUME_2) {
        return _dispatch_release_2_tailcall(dq);
    }
}

依次下符号断点走了_dispatch_runloop_queue_poke逻辑：

static void
_dispatch_runloop_queue_poke(dispatch_lane_t dq, dispatch_qos_t qos,
        dispatch_wakeup_flags_t flags)
{
    ……

    if (dx_type(dq) == DISPATCH_QUEUE_MAIN_TYPE) {
        dispatch_once_f(&_dispatch_main_q_handle_pred, dq,
                _dispatch_runloop_queue_handle_init);
    }
    ……
}

核心在_dispatch_runloop_queue_handle_init中的实现：

static void
_dispatch_runloop_queue_handle_init(void *ctxt)
{
    dispatch_lane_t dq = (dispatch_lane_t)ctxt;
    //创建handle
    dispatch_runloop_handle_t handle;

    _dispatch_fork_becomes_unsafe();

#if TARGET_OS_MAC
    mach_port_options_t opts = {
        .flags = MPO_CONTEXT_AS_GUARD | MPO_STRICT | MPO_INSERT_SEND_RIGHT,
    };
    mach_port_context_t guard = (uintptr_t)dq;
    kern_return_t kr;
    mach_port_t mp;

    if (dx_type(dq) == DISPATCH_QUEUE_MAIN_TYPE) {
        opts.flags |= MPO_QLIMIT;
        opts.mpl.mpl_qlimit = 1;
    }

    kr = mach_port_construct(mach_task_self(), &opts, guard, &mp);
    DISPATCH_VERIFY_MIG(kr);
    (void)dispatch_assume_zero(kr);

    handle = mp;
#elif defined(__linux__)
……
#else
#error "runloop support not implemented on this platform"
#endif
    _dispatch_runloop_queue_set_handle(dq, handle);

    _dispatch_program_is_probably_callback_driven = true;
}

首先创建handle。然后handle = mp进行赋值。
_dispatch_runloop_queue_set_handle将handle关联到dq的do_ctxt中。

static inline dispatch_runloop_handle_t
_dispatch_runloop_queue_get_handle(dispatch_lane_t dq)
{
#if TARGET_OS_MAC
    return ((dispatch_runloop_handle_t)(uintptr_t)dq->do_ctxt);
#elif defined(__linux__)
    // decode: 0 is a valid fd, so offset by 1 to distinguish from NULL
    return ((dispatch_runloop_handle_t)(uintptr_t)dq->do_ctxt) - 1;
#elif defined(_WIN32)
    return ((dispatch_runloop_handle_t)(uintptr_t)dq->do_ctxt);
#else
#error "runloop support not implemented on this platform"
#endif
}

在block内部断点有如下调用栈：

* thread #1, queue = 'com.apple.main-thread', stop reason = breakpoint 1.1
  * frame #0: 0x0000000102fef477 GCDDemo`__29-[ViewController viewDidLoad]_block_invoke(.block_descriptor=0x0000000102ff2040) at ViewController.m:33:9
    frame #1: 0x00000001032647ec libdispatch.dylib`_dispatch_call_block_and_release + 12
    frame #2: 0x00000001032659c8 libdispatch.dylib`_dispatch_client_callout + 8
    frame #3: 0x0000000103273e75 libdispatch.dylib`_dispatch_main_queue_callback_4CF + 1152
    frame #4: 0x00007fff2038fdbb CoreFoundation`__CFRUNLOOP_IS_SERVICING_THE_MAIN_DISPATCH_QUEUE__ + 9
    frame #5: 0x00007fff2038a63e CoreFoundation`__CFRunLoopRun + 2685

CoreFoundation调用的回调。_dispatch_main_queue_callback_4CF实现：

void
_dispatch_main_queue_callback_4CF(
        void *ignored DISPATCH_UNUSED)
{
    // the main queue cannot be suspended and no-one looks at this bit
    // so abuse it to avoid dirtying more memory

    if (_dispatch_main_q.dq_side_suspend_cnt) {
        return;
    }
    _dispatch_main_q.dq_side_suspend_cnt = true;
    _dispatch_main_queue_drain(&_dispatch_main_q);
    _dispatch_main_q.dq_side_suspend_cnt = false;
}

调用了_dispatch_main_queue_drain：

image.png
到这里就与全局队列调用相同的逻辑了。

但是有个问题是_dispatch_main_queue_callback_4CF是谁传递给CF的呢？在CoreFoundation动态库中搜索发现是写死的：

image.png

6.2.4 _dispatch_lane_push（自己创建串行队列）

dispatch_queue_t queue = dispatch_queue_create("test", NULL);
dispatch_async(queue, ^{
    NSLog(@"test");
});

同样进入_dispatch_lane_push逻辑：

void
_dispatch_lane_push(dispatch_lane_t dq, dispatch_object_t dou,
        dispatch_qos_t qos)
{
    dispatch_wakeup_flags_t flags = 0;
    struct dispatch_object_s *prev;

    if (unlikely(_dispatch_object_is_waiter(dou))) {
        return _dispatch_lane_push_waiter(dq, dou._dsc, qos);
    }

    dispatch_assert(!_dispatch_object_is_global(dq));
    qos = _dispatch_queue_push_qos(dq, qos);

    ……
    os_mpsc_push_update_prev(os_mpsc(dq, dq_items), prev, dou._do, do_next);
    if (flags) {
        //_dispatch_lane_wakeup
        return dx_wakeup(dq, qos, flags);
    }
}

断点跟踪会进入_dispatch_lane_wakeup逻辑：

void
_dispatch_lane_wakeup(dispatch_lane_class_t dqu, dispatch_qos_t qos,
        dispatch_wakeup_flags_t flags)
{
    dispatch_queue_wakeup_target_t target = DISPATCH_QUEUE_WAKEUP_NONE;

    if (unlikely(flags & DISPATCH_WAKEUP_BARRIER_COMPLETE)) {
   //有barrier走这里
        return _dispatch_lane_barrier_complete(dqu, qos, flags);
    }
    if (_dispatch_queue_class_probe(dqu)) {
        target = DISPATCH_QUEUE_WAKEUP_TARGET;
    }
//否则走这里
    return _dispatch_queue_wakeup(dqu, qos, flags, target);
}

继续断点会进入_dispatch_queue_wakeup逻辑：

image.png

这个时候断点并没有断住。block内部打断点堆栈调用如下：

* thread #7, queue = 'test', stop reason = breakpoint 1.1
  * frame #0: 0x0000000103765477 GCDDemo`__29-[ViewController viewDidLoad]_block_invoke(.block_descriptor=0x0000000103768038) at ViewController.m:34:9
    frame #1: 0x00000001039da7ec libdispatch.dylib`_dispatch_call_block_and_release + 12
    frame #2: 0x00000001039db9c8 libdispatch.dylib`_dispatch_client_callout + 8
    frame #3: 0x00000001039e2296 libdispatch.dylib`_dispatch_lane_serial_drain + 796
    frame #4: 0x00000001039e2f67 libdispatch.dylib`_dispatch_lane_invoke + 439
    frame #5: 0x00000001039eede2 libdispatch.dylib`_dispatch_workloop_worker_thread + 882
    frame #6: 0x00007fff61163a3d libsystem_pthread.dylib`_pthread_wqthread + 290
    frame #7: 0x00007fff61162b77 libsystem_pthread.dylib`start_wqthread + 15

_dispatch_workloop_worker_thread是在_dispatch_root_queues_init_once中赋值的：

image.png
但是断点并没有进入这里，直接添加断点启动App发现是在main之后调用赋值的：

(lldb) bt
* thread #1, queue = 'com.apple.main-thread', stop reason = breakpoint 3.1
  * frame #0: 0x000000010bb53d0e libdispatch.dylib`_dispatch_root_queues_init_once
    frame #1: 0x000000010bb419c8 libdispatch.dylib`_dispatch_client_callout + 8
    frame #2: 0x000000010bb42f33 libdispatch.dylib`_dispatch_once_callout + 66
    frame #3: 0x000000010bb4e5c3 libdispatch.dylib`_dispatch_root_queue_poke_slow + 363
    frame #4: 0x00007fff2467a0a1 UIKitCore`_UIApplicationMainPreparations + 91
    frame #5: 0x00007fff2467a01c UIKitCore`UIApplicationMain + 73
    frame #6: 0x000000010b8cbf72 GCDDemo`main(argc=1, argv=0x00007ffee4333cd8) at main.m:17:12
    frame #7: 0x00007fff20256409 libdyld.dylib`start + 1
    frame #8: 0x00007fff20256409 libdyld.dylib`start + 1

UIKitCore中调用_dispatch_root_queue_poke_slow：

static void
_dispatch_root_queue_poke_slow(dispatch_queue_global_t dq, int n, int floor)
{
    ......

    _dispatch_root_queues_init();
    _dispatch_debug_root_queue(dq, __func__);
    _dispatch_trace_runtime_event(worker_request, dq, (uint64_t)n);
    ......
}

核心是_dispatch_root_queues_init：

static inline void
_dispatch_root_queues_init(void)
{
    dispatch_once_f(&_dispatch_root_queues_pred, NULL,
            _dispatch_root_queues_init_once);
}

这样就调用到了_dispatch_root_queues_init_once。

6.2.5 _dispatch_lane_concurrent_push（自己创建并行队列）

dispatch_queue_t queue = dispatch_queue_create("test", DISPATCH_QUEUE_CONCURRENT);
dispatch_async(queue, ^{
    NSLog(@"test");
});

同样会进入_dispatch_lane_concurrent_push流程：

void
_dispatch_lane_concurrent_push(dispatch_lane_t dq, dispatch_object_t dou,
        dispatch_qos_t qos)
{
    // <rdar://problem/24738102&24743140> reserving non barrier width
    // doesn't fail if only the ENQUEUED bit is set (unlike its barrier
    // width equivalent), so we have to check that this thread hasn't
    // enqueued anything ahead of this call or we can break ordering
    if (dq->dq_items_tail == NULL &&
            !_dispatch_object_is_waiter(dou) &&
            !_dispatch_object_is_barrier(dou) &&
            _dispatch_queue_try_acquire_async(dq)) {
        return _dispatch_continuation_redirect_push(dq, dou, qos);
    }

    _dispatch_lane_push(dq, dou, qos);
}

这里有_dispatch_lane_push也就是说某些情况下会进入串行队列的逻辑。有barrier那么意味着在有栅栏的情况下有可能会进入_dispatch_lane_push的逻辑，栅栏部分将在后续文章分析。

会进入_dispatch_continuation_redirect_push流程：

static void
_dispatch_continuation_redirect_push(dispatch_lane_t dl,
        dispatch_object_t dou, dispatch_qos_t qos)
{
    if (likely(!_dispatch_object_is_redirection(dou))) {
        dou._dc = _dispatch_async_redirect_wrap(dl, dou);
    } else if (!dou._dc->dc_ctxt) {
        // find first queue in descending target queue order that has
        // an autorelease frequency set, and use that as the frequency for
        // this continuation.
        dou._dc->dc_ctxt = (void *)
        (uintptr_t)_dispatch_queue_autorelease_frequency(dl);
    }
       //target发生了变化
    dispatch_queue_t dq = dl->do_targetq;
    if (!qos) qos = _dispatch_priority_qos(dq->dq_priority);
    dx_push(dq, dou, qos);
}

do_targetq赋值给了dq。此时dq_push已经不是_dispatch_lane_concurrent_push而是queue_pthread_root类型的_dispatch_root_queue_push。这个时候逻辑就和全局并发队列一致了（经验证内部调用流程逻辑也一致）。

do_targetq是在_dispatch_lane_create_with_target中赋值的tq：

image.png

block内部断点堆栈如下：

* thread #3, queue = 'test', stop reason = breakpoint 2.1
  * frame #0: 0x0000000101ca7477 GCDDemo`__29-[ViewController viewDidLoad]_block_invoke(.block_descriptor=0x0000000101caa040) at ViewController.m:34:9
    frame #1: 0x0000000101f1c7ec libdispatch.dylib`_dispatch_call_block_and_release + 12
    frame #2: 0x0000000101f1d9c8 libdispatch.dylib`_dispatch_client_callout + 8
    frame #3: 0x0000000101f20316 libdispatch.dylib`_dispatch_continuation_pop + 557
    frame #4: 0x0000000101f1f71c libdispatch.dylib`_dispatch_async_redirect_invoke + 779
    frame #5: 0x0000000101f2f508 libdispatch.dylib`_dispatch_root_queue_drain + 351
    frame #6: 0x0000000101f2fe6d libdispatch.dylib`_dispatch_worker_thread2 + 135
    frame #7: 0x00007fff611639f7 libsystem_pthread.dylib`_pthread_wqthread + 220
    frame #8: 0x00007fff61162b77 libsystem_pthread.dylib`start_wqthread + 15

整个调用流程如下：

同步异步函数调用流程图

七、案例分析

7.1 案例1

@property (atomic, assign) int num;

- (void)test {
    while (self.num < 5) {
        dispatch_async(dispatch_get_global_queue(0, 0), ^{
            self.num++;
        });
    }
    NSLog(@"result: %d",self.num);
}

while循环结束的条件是num >= 5（这里就确定了至少会是5）。
num++的逻辑是在异步并发队列执行的，不会阻塞流程。
由于判断条件是num < 5的时候进入循环执行逻辑，所以最终会存在num++还没有执行完毕while循环继续进入执行，这样num最终就>=5了。本质上是任务执行次数>=5次。

将dispatch_async改为同步函数dispatch_sync的时候就和普通的循环没有区别了，最终num会为5。

将队列改为自己创建的全局串行队列仍然不解决问题，只是顺序执行仍然会进入>=5次。

将队列改为异步main函数会造成while死循环，任务一次都不执行，因为任务本身要等while执行完毕。

既然是异步执行的，那么NSLog输出的num是最终的值么？不一定，因为输出的时候任务不一定全部执行完毕，可以加个延时测试下：

- (void)test {
    while (self.num < 5) {
        dispatch_async(dispatch_get_global_queue(0, 0), ^{
            self.num++;
        });
    }
    NSLog(@"result: %d",self.num);
    dispatch_after(dispatch_time(DISPATCH_TIME_NOW, (int64_t)(3 * NSEC_PER_SEC)), dispatch_get_main_queue(), ^{
        NSLog(@"after result: %d",self.num);
    });
}

输出:

result: 7
after result: 16

7.2 案例2

@property (atomic, assign) int num;

- (void)test {
    for (int i = 0; i < 10000; i++) {
        dispatch_async(dispatch_get_global_queue(0, 0), ^{
            self.num++;
        });
    }
    NSLog(@"result: %d",self.num);
}

循环结束条件是i = 10000（这里就控制了循环只执行10000次）。
num++是在异步并发队列执行的。
由于任务是异步执行的，所以result输出<= 10000（有任务还没有回来）。

同理加个延迟之后num的值是否就为10000了？

- (void)test {
    for (int i = 0; i < 10000; i++) {
        dispatch_async(dispatch_get_global_queue(0, 0), ^{
            self.num++;
        });
    }
    NSLog(@"result: %d",self.num);
    dispatch_after(dispatch_time(DISPATCH_TIME_NOW, (int64_t)(10 * NSEC_PER_SEC)), dispatch_get_main_queue(), ^{
        NSLog(@"after result: %d",self.num);
    });
}

输出：

result: 9956
after result: 9964

为什么？因为任务虽然执行了10000次，由于并发block获取的num多次有可能相同，这就导致了值是小于10000的。

将dispatch_async改为同步函数dispatch_sync就和普通的循环没有区别了，最终num会为10000。
将队列改为自己创建的全局串行队列这个时候最终num = 10000（由于串行队列顺序执行）

- (void)test {
    dispatch_queue_t queue = dispatch_queue_create("test", NULL);
    for (int i = 0; i < 10000; i++) {
        dispatch_async(queue, ^{
            self.num++;
        });
    }
    NSLog(@"result: %d",self.num);
    dispatch_after(dispatch_time(DISPATCH_TIME_NOW, (int64_t)(10 * NSEC_PER_SEC)), dispatch_get_main_queue(), ^{
        NSLog(@"after result: %d",self.num);
    });
}

输出：

result: 3049
after result: 10000

将队列改为异步main函数:

- (void)test {
    for (int i = 0; i < 10000; i++) {
        dispatch_async(dispatch_get_main_queue(), ^{
            self.num++;
        });
    }
    NSLog(@"result: %d",self.num);
    dispatch_after(dispatch_time(DISPATCH_TIME_NOW, (int64_t)(10 * NSEC_PER_SEC)), dispatch_get_main_queue(), ^{
        NSLog(@"after result: %d",self.num);
    });
}

输出：

result: 0
after result: 10000

由于都在main队列，所以会先执行完for循环，然后输出result，再执行任务。
所以先输出0，延时后输出10000。

⚠️atomic只能保证setter & getter内部的安全，多次settter之间并不是安全的。

GCD底层分析（一）：队列、同步异步函数

一、GCD 简介

1.1 GCD

1.2 GCD 的作用

二、函数与队列

2.1 同步与异步

2.2 串行队列与并发队列

2.2.1 案例1

2.2.2 案例2

2.2.3 案例3

2.2.4 案例4

三、主队列与全局并发队列

3.1 主队列

3.1.1 dispatch_get_main_queue 源码分析

3.2 全局队列

四、dispatch_queue_create

4.1 _dispatch_lane_create_with_target

4.2 _dispatch_queue_attr_to_info

4.3 _dispatch_object_alloc

4.3.1 _os_object_alloc_realized

4.4 _dispatch_queue_init

4.5 _dispatch_trace_queue_create

五、dispatch_queue_t

5.1 DISPATCH_DECL 源码分析

5.2 dispatch_queue_s 分析

5.3 dispatch_object_s 分析

六、GCD 函数调用流程

6.1 同步流程

6.1.1 dispatch_sync 源码分析

6.1.2 _dispatch_sync_f_inline

6.1.3 _dispatch_barrier_sync_f（串行）

6.1.4 _dispatch_sync_f_slow（全局并发）

6.1.5 _dispatch_sync_invoke_and_complete（自定义并发）

6.2 异步流程

6.2.1 dispatch_async 源码分析

6.2.2 _dispatch_continuation_async

6.2.2 _dispatch_root_queue_push（全局队列）

6.2.3 _dispatch_main_queue_push（主队列）

6.2.4 _dispatch_lane_push（自己创建串行队列）

6.2.5 _dispatch_lane_concurrent_push（自己创建并行队列）

七、案例分析

7.1 案例1

7.2 案例2

猜你喜欢

热点阅读