讀寫鎖的概念很簡單，允許多個線程同時獲取讀鎖，但同一時間只允許一個線程獲得寫鎖，因此也稱作共享-獨占鎖。在C#中，推薦使用ReaderWriterLockSlim類來完成讀寫鎖的功能。
某些場合下，對一個對象的讀取次數遠遠大於修改次數，如果只是簡單的用lock方式加鎖，則會影響讀取的效率。而如果采用讀寫鎖，則多個線程可以同時讀取該對象，只有等到對象被寫入鎖占用的時候，才會阻塞。
簡單的說，當某個線程進入讀取模式時，此時其他線程依然能進入讀取模式，假設此時一個線程要進入寫入模式，那麼他不得不被阻塞。直到讀取模式退出為止。
C#多線程編程實例 線程與窗體交互【附源碼】 http://www.linuxidc.com/Linux/2014-07/104294.htm
C#數學運算表達式解釋器 http://www.linuxidc.com/Linux/2014-07/104289.htm
在C語言中解析JSON配置文件 http://www.linuxidc.com/Linux/2014-05/101822.htm
C++ Primer Plus 第6版 中文版 清晰有書簽PDF+源代碼 http://www.linuxidc.com/Linux/2014-05/101227.htm
同樣的，如果某個線程進入了寫入模式，那麼其他線程無論是要寫入還是讀取，都是會被阻塞的。
進入寫入/讀取模式有2種方法：
EnterReadLock嘗試進入寫入模式鎖定狀態。
TryEnterReadLock(Int32) 嘗試進入讀取模式鎖定狀態，可以選擇整數超時時間。
EnterWriteLock 嘗試進入寫入模式鎖定狀態。
TryEnterWriteLock(Int32) 嘗試進入寫入模式鎖定狀態，可以選擇超時時間。
退出寫入/讀取模式有2種方法：
ExitReadLock 減少讀取模式的遞歸計數，並在生成的計數為 0（零）時退出讀取模式。
ExitWriteLock 減少寫入模式的遞歸計數，並在生成的計數為 0（零）時退出寫入模式。
下面演示一下用法：
public class Program
{
static private ReaderWriterLockSlim rwl = new ReaderWriterLockSlim();
static void Main(string[] args)
{
Thread t_read1 = new Thread(new ThreadStart(ReadSomething));
t_read1.Start();
Console.WriteLine("{0} Create Thread ID {1} , Start ReadSomething", DateTime.Now.ToString("hh:mm:ss fff"), t_read1.GetHashCode());
Thread t_read2 = new Thread(new ThreadStart(ReadSomething));
t_read2.Start();
Console.WriteLine("{0} Create Thread ID {1} , Start ReadSomething", DateTime.Now.ToString("hh:mm:ss fff"), t_read2.GetHashCode());
Thread t_write1 = new Thread(new ThreadStart(WriteSomething));
t_write1.Start();
Console.WriteLine("{0} Create Thread ID {1} , Start WriteSomething", DateTime.Now.ToString("hh:mm:ss fff"), t_write1.GetHashCode());
}
static public void ReadSomething()
{
Console.WriteLine("{0} Thread ID {1} Begin EnterReadLock...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
rwl.EnterReadLock();
try
{
Console.WriteLine("{0} Thread ID {1} reading sth...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
Thread.Sleep(5000);//模擬讀取信息
Console.WriteLine("{0} Thread ID {1} reading end.", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
}
finally
{
rwl.ExitReadLock();
Console.WriteLine("{0} Thread ID {1} ExitReadLock...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
}
}
static public void WriteSomething()
{
Console.WriteLine("{0} Thread ID {1} Begin EnterWriteLock...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
rwl.EnterWriteLock();
try
{
Console.WriteLine("{0} Thread ID {1} writing sth...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
Thread.Sleep(10000);//模擬寫入信息
Console.WriteLine("{0} Thread ID {1} writing end.", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
}
finally
{
rwl.ExitWriteLock();
Console.WriteLine("{0} Thread ID {1} ExitWriteLock...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
}
}
}
可以看到3號線程和4號線程能夠同時進入讀模式，而5號線程過了5秒鐘後（即3,4號線程退出讀鎖後），才能進入寫模式。
把上述代碼修改一下，先開啟2個寫模式的線程，然後在開啟讀模式線程，代碼如下：
{
Thread t_write1 = new Thread(new ThreadStart(WriteSomething));
t_write1.Start();
Console.WriteLine("{0} Create Thread ID {1} , Start WriteSomething", DateTime.Now.ToString("hh:mm:ss fff"), t_write1.GetHashCode());
Thread t_write2 = new Thread(new ThreadStart(WriteSomething));
t_write2.Start();
Console.WriteLine("{0} Create Thread ID {1} , Start WriteSomething", DateTime.Now.ToString("hh:mm:ss fff"), t_write2.GetHashCode());
Thread t_read1 = new Thread(new ThreadStart(ReadSomething));
t_read1.Start();
Console.WriteLine("{0} Create Thread ID {1} , Start ReadSomething", DateTime.Now.ToString("hh:mm:ss fff"), t_read1.GetHashCode());
Thread t_read2 = new Thread(new ThreadStart(ReadSomething));
t_read2.Start();
Console.WriteLine("{0} Create Thread ID {1} , Start ReadSomething", DateTime.Now.ToString("hh:mm:ss fff"), t_read2.GetHashCode());
}
結果如下：
可以看到，3號線程和4號線程都要進入寫模式，但是3號線程先占用寫入鎖，因此4號線程不得不等了10s後才進入。5號線程和6號線程需要占用讀取鎖，因此等4號線程退出寫入鎖後才能繼續下去。
TryEnterReadLock和TryEnterWriteLock可以設置一個超時時間，運行到這句話的時候，線程會阻塞在此，如果此時能占用鎖，那麼返回true，如果到超時時間還未占用鎖，那麼返回false，放棄鎖的占用，直接繼續執行下面的代碼。
EnterUpgradeableReadLock
ReaderWriterLockSlim類提供了可升級讀模式，這種方式和讀模式的區別在於它還有通過調用 EnterWriteLock 或 TryEnterWriteLock 方法升級為寫入模式。 因為每次只能有一個線程處於可升級模式。進入可升級模式的線程，不會影響讀取模式的線程，即當一個線程進入可升級模式，任意數量線程可以同時進入讀取模式，不會阻塞。如果有多個線程已經在等待獲取寫入鎖，那麼運行EnterUpgradeableReadLock將會阻塞，直到那些線程超時或者退出寫入鎖。
下面代碼演示了如何在可升級讀模式下，升級到寫入鎖。
static public void UpgradeableRead()
{
Console.WriteLine("{0} Thread ID {1} Begin EnterUpgradeableReadLock...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
rwl.EnterUpgradeableReadLock();
try
{
Console.WriteLine("{0} Thread ID {1} doing sth...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
Console.WriteLine("{0} Thread ID {1} Begin EnterWriteLock...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
rwl.EnterWriteLock();
try
{
Console.WriteLine("{0} Thread ID {1} writing sth...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
Thread.Sleep(10000);//模擬寫入信息
Console.WriteLine("{0} Thread ID {1} writing end.", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
}
finally
{
rwl.ExitWriteLock();
Console.WriteLine("{0} Thread ID {1} ExitWriteLock...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
}
Thread.Sleep(10000);//模擬讀取信息
Console.WriteLine("{0} Thread ID {1} doing end.", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
}
finally
{
rwl.ExitUpgradeableReadLock();
Console.WriteLine("{0} Thread ID {1} ExitUpgradeableReadLock...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
}
}
讀寫鎖對於性能的影響是明顯的。
下面測試代碼：
public class Program
{
static private ReaderWriterLockSlim rwl = new ReaderWriterLockSlim();
static void Main(string[] args)
{
Stopwatch sw = new Stopwatch();
sw.Start();
List<Task> lstTask = new List<Task>();
for (int i = 0; i < 500; i++)
{
if (i % 25 != 0)
{
var t = Task.Factory.StartNew(ReadSomething);
lstTask.Add(t);
}
else
{
var t = Task.Factory.StartNew(WriteSomething);
lstTask.Add(t);
}
}
Task.WaitAll(lstTask.ToArray());
sw.Stop();
Console.WriteLine("使用ReaderWriterLockSlim方式，耗時：" + sw.Elapsed);
sw.Restart();
lstTask = new List<Task>();
for (int i = 0; i < 500; i++)
{
if (i % 25 != 0)
{
var t = Task.Factory.StartNew(ReadSomething_lock);
lstTask.Add(t);
}
else
{
var t = Task.Factory.StartNew(WriteSomething_lock);
lstTask.Add(t);
}
}
Task.WaitAll(lstTask.ToArray());
sw.Stop();
Console.WriteLine("使用lock方式，耗時：" + sw.Elapsed);
}
static private object _lock1 = new object();
static public void ReadSomething_lock()
{
lock (_lock1)
{
//Console.WriteLine("{0} Thread ID {1} reading sth...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
Thread.Sleep(10);//模擬讀取信息
//Console.WriteLine("{0} Thread ID {1} reading end.", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
}
}
static public void WriteSomething_lock()
{
lock (_lock1)
{
//Console.WriteLine("{0} Thread ID {1} writing sth...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
Thread.Sleep(100);//模擬寫入信息
//Console.WriteLine("{0} Thread ID {1} writing end.", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
}
}
static public void ReadSomething()
{
rwl.EnterReadLock();
try
{
//Console.WriteLine("{0} Thread ID {1} reading sth...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
Thread.Sleep(10);//模擬讀取信息
//Console.WriteLine("{0} Thread ID {1} reading end.", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
}
finally
{
rwl.ExitReadLock();
}
}
static public void WriteSomething()
{
rwl.EnterWriteLock();
try
{
//Console.WriteLine("{0} Thread ID {1} writing sth...", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
Thread.Sleep(100);//模擬寫入信息
//Console.WriteLine("{0} Thread ID {1} writing end.", DateTime.Now.ToString("hh:mm:ss fff"), Thread.CurrentThread.GetHashCode());
}
finally
{
rwl.ExitWriteLock();
}
}
}
上述代碼，就500個Task，每個Task占用一個線程池線程，其中20個寫入線程和480個讀取線程，模擬操作。其中讀取數據花10ms，寫入操作花100ms，分別測試了對於lock方式和ReaderWriterLockSlim方式。可以做一個估算，對於ReaderWriterLockSlim，假設480個線程同時讀取，那麼消耗10ms，20個寫入操作占用2000ms，因此所消耗時間2010ms，而對於普通的lock方式，由於都是獨占性的，因此480個讀取操作占時間4800ms+20個寫入操作2000ms=6800ms。運行結果顯示了性能提升明顯。
