首页 > 代码库 > C#高级编程五十八天----并行集合
C#高级编程五十八天----并行集合
并行集合
对于并行任务,与其相关紧密的就是对一些共享资源,数据结构的并行訪问.常常要做的就是对一些队列进行加锁-解锁,然后运行类似插入,删除等等相互排斥操作. .NET4提供了一些封装好的支持并行操作数据容器,能够降低并行编程的复杂程度.
并行集合的命名空间:System.Collections.Concurrent
并行容器:
ConcurrentQueue
ConcurrentStack
ConcurrentBag: 一个无序的数据结构集,当不考虑顺序时很实用.
BlockingCollection:与经典的堵塞队列数据结构类似
ConcurrentDictoinary
以上这些集合在某种程度上使用了无锁技术(CAS和内存屏蔽),与加相互排斥锁相比获得了性能的提升.可是在串行程序中,最好不要使用这些集合,他们必定会影响性能.
ConcurrentQueue使用方法与实例
其全然无锁,但当CAS面临资源竞争失败时可能会陷入自旋并重试操作.
Enqueue:在队尾插入元素
TryDequeue:尝试删除对头元素,并通过out參数返回
TryPeek:尝试将对头元素通过out參数返回,但不删除元素
案例:
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using System.Collections.Concurrent;
namespace 并行结合Queue
{
class Program
{
internal static ConcurrentQueue<int> _TestQueue;
class ThreadWork1 //生产者
{
public ThreadWork1()
{ }
public void run()
{
Console.WriteLine("ThreadWork1 run { ");
for (int i = 0; i < 100; i++)
{
Console.WriteLine("ThreadWork1 producer: "+i);
_TestQueue.Enqueue(i);
}
Console.WriteLine("ThreadWork1 run } ");
}
}
class ThreadWork2 //consumer
{
public ThreadWork2()
{ }
public void run()
{
int i = 0;
bool IsDequeue = false;
Console.WriteLine("ThreadWork2 run { ");
for (; ; )
{
IsDequeue = _TestQueue.TryDequeue(out i);
if (IsDequeue)
{
System.Console.WriteLine("ThreadWork2 consumer: " + i * i + " =====");
}
if (i==99)
{
break;
}
}
Console.WriteLine("ThreadWork2 run } ");
}
}
static void StartT1()
{
ThreadWork1 work1 = new ThreadWork1();
work1.run();
}
static void StartT2()
{
ThreadWork2 work2 = new ThreadWork2();
work2.run();
}
static void Main(string[] args)
{
Task t1 = new Task(() => StartT1());
Task t2 = new Task(() => StartT2());
_TestQueue = new ConcurrentQueue<int>();
Console.WriteLine("Sample 3-1 Main {");
Console.WriteLine("Main t1 t2 started {");
t1.Start();
t2.Start();
Console.WriteLine("Main t1 t2 started }");
Console.WriteLine("Main wait t1 t2 end {");
Task.WaitAll(t1, t2);
Console.WriteLine("Main wait t1 t2 end }");
Console.WriteLine("Sample 3-1 Main }");
Console.ReadKey();
}
}
}
ConcurrentStact
其全然无锁,但当CAS面临资源竞争失败时可能会陷入自旋并重试操作.
Push:向栈顶插入元素
TryPop:从栈顶弹出元素,而且通过out參数返回
TryPeek:返回栈顶元素,但不弹出
案例:
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using System.Collections.Concurrent;
namespace 并行结合Queue
{
class Program
{
internal static ConcurrentStack<int> _TestStack;
class ThreadWork1 //生产者
{
public ThreadWork1()
{ }
public void run()
{
Console.WriteLine("ThreadWork1 run { ");
for (int i = 0; i < 100; i++)
{
Console.WriteLine("ThreadWork1 producer: "+i);
_TestStack.Push(i);
}
Console.WriteLine("ThreadWork1 run } ");
}
}
class ThreadWork2 //consumer
{
public ThreadWork2()
{ }
public void run()
{
int i = 0;
bool IsDequeue = false;
Console.WriteLine("ThreadWork2 run { ");
for (; ; )
{
IsDequeue = _TestStack.TryPop(out i);
if (IsDequeue)
{
System.Console.WriteLine("ThreadWork2 consumer: " + i * i + " =====" + i);
}
if (i==99)
{
break;
}
}
Console.WriteLine("ThreadWork2 run } ");
}
}
static void StartT1()
{
ThreadWork1 work1 = new ThreadWork1();
work1.run();
}
static void StartT2()
{
ThreadWork2 work2 = new ThreadWork2();
work2.run();
}
static void Main(string[] args)
{
Task t1 = new Task(() => StartT1());
Task t2 = new Task(() => StartT2());
_TestStack = new ConcurrentStack<int>();
Console.WriteLine("Sample 4-1 Main {");
Console.WriteLine("Main t1 t2 started {");
t1.Start();
t2.Start();
Console.WriteLine("Main t1 t2 started }");
Console.WriteLine("Main wait t1 t2 end {");
Task.WaitAll(t1, t2);
Console.WriteLine("Main wait t1 t2 end }");
Console.WriteLine("Sample 4-1 Main }");
Console.ReadKey();
}
}
}
ConcurrentBag
一个无序的集合,程序能够向当中插入元素,或删除元素.
在同一个线程中向集合插入,删除元素效率非常高.
Add:向集合中插入元素
TryTake:从集合中取出元素并删除
TryPeek:从集合中取出元素,但不删除元素
案例:
using System;
using System.Collections.Concurrent;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
namespace 并行集合ConcurrentBag
{
class Program
{
internal static ConcurrentBag<int> _TestBag;
class ThreadWork1 //producer
{
public ThreadWork1()
{ }
public void run()
{
Console.WriteLine("Threadwork1 run { ");
for (int i = 0; i < 100; i++)
{
Console.WriteLine("ThreadWork1 producer: "+i);
_TestBag.Add(i);
}
Console.WriteLine("ThreadWork1 run } ");
}
}
class ThreadWork2//consumer
{
public ThreadWork2()
{ }
public void run()
{
bool IsDequeue = false;
Console.WriteLine("ThreadWork2 run { ");
for (int i = 0; i < 100; i++)
{
IsDequeue = _TestBag.TryTake(out i);
if (IsDequeue)
{
Console.WriteLine("ThreadWork2 consumer: " + i * i + "======" + i);
}
}
Console.WriteLine("ThreadWork2 run } ");
}
}
static void Start1()
{
ThreadWork1 work1 = new ThreadWork1();
work1.run();
}
static void Start2()
{
ThreadWork2 work2 = new ThreadWork2();
work2.run();
}
static void Main(string[] args)
{
Task t1 = new Task(() => Start1());
Task t2 = new Task(() => Start2());
_TestBag = new ConcurrentBag<int>();
t1.Start();
t2.Start();
Console.WriteLine("Main t1 t2 started }");
Console.WriteLine("Main wait t1 t2 end {");
Task.WaitAll(t1, t2);
Console.WriteLine("Main wait t1 t2 end }");
Console.WriteLine("Sample 4-3 Main }");
Console.ReadKey();
}
}
}
BlockingCollection
一个支持界限和堵塞的容器
Add:向容器中插入元素
TryTake:从容器中取出元素并删除
TryPeek:从容器中取出元素,但不删除
CompleteAdding:告诉容器,加入元素完毕.此时假设还想继续加入会发生异常.
IsCompleted:告诉消费者线程,产生者线程还在继续执行中,任务还未完毕.
案例:
程序中,消费者线程全然使用While(!__testCollection.IsCompleted)作为退出执行的推断条件.在Work1中,有两条语句被凝视了,当i为50时设置为CompleteAdding,但当继续向当中插入元素时,系统抛出异常,提示无法再继续插入.
案例:
using System;
using System.Collections.Concurrent;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
namespace 并行结合Collection
{
class Program
{
internal static BlockingCollection<int> _testBCollection;
class ThreadWork1 //producer
{
public ThreadWork1()
{ }
public void run()
{
Console.WriteLine("ThreadWork1 run { ");
for (int i = 0; i < 100; i++)
{
Console.WriteLine("ThreadWork1 producer: "+i);
_testBCollection.Add(i);
//当i为50时设置为CompleteAdding,但当继续向当中插入元素时,系统抛出异常,提示无法再继续插入.
//if (i==50)
//{
// _testBCollection.CompleteAdding();
//}
}
_testBCollection.CompleteAdding();
Console.WriteLine("ThreadWork1 run } ");
}
}
class ThreadWork2//consumer
{
public ThreadWork2()
{ }
public void run()
{
int i = 0;
bool IsDequeue = false;
Console.WriteLine("ThreadWork2 run { ");
while (!_testBCollection.IsCompleted)
{
//不明确这里为啥i会自己主动的++
//Console.WriteLine("i=================="+i);
IsDequeue = _testBCollection.TryTake(out i);
if (IsDequeue)
{
Console.WriteLine("ThreadWork2 consumer: "+i*i+"====="+i);
//i++;这句话不加还是会出现自己主动++的操作
}
}
}
}
static void StartT1()
{
ThreadWork1 work1 = new ThreadWork1();
work1.run();
}
static void StartT2()
{
ThreadWork2 work2 = new ThreadWork2();
work2.run();
}
static void Main(string[] args)
{
Task t1 = new Task(() => StartT1());
Task t2 = new Task(() => StartT2());
_testBCollection = new BlockingCollection<int>();
Console.WriteLine("Sample 4-4 Main {");
Console.WriteLine("Main t1 t2 started {");
t1.Start();
t2.Start();
Console.WriteLine("Main t1 t2 started }");
Console.WriteLine("Main wait t1 t2 end {");
Task.WaitAll(t1, t2);
Console.WriteLine("Main wait t1 t2 end }");
Console.WriteLine("Sample 4-4 Main }");
Console.ReadKey();
}
}
}
分析://_testBCollection.CompleteAdding();//这句话凝视掉work2陷入循环,无法退出
ConcurrentDictionary
对于读操作是全然无锁的,当非常多线程要改动数据时,它会使用细粒度的锁.
AddOrUpdate:假设键不存在,方法会在容器中加入新的键和值,假设存在,则更新现有的键和值
GetOrAdd:假设键不存在,方法会向容器中加入新的键和值,假设存在则返回现有的值,并不加入新值.
TryAdd:尝试在容器中加入新的键和值
TryGetValue:尝试依据指定的键获得值
TryUpdate:有条件的更新当前键所相应的值
TryRemove:尝试删除指定的键.
Getenumerator:返回一个可以遍历整个容器的枚举器.
案例:
using System;
using System.Collections.Concurrent;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
namespace 并行集合Dictionary
{
class Program
{
internal static ConcurrentDictionary<int, int> _TestDictionary;
class ThreadWork1//producer
{
public ThreadWork1()
{ }
public void run()
{
System.Console.WriteLine("ThreadWork1 run { ");
for (int i = 0; i < 100; i++)
{
System.Console.WriteLine("ThreadWork1 producer: " + i);
_TestDictionary.TryAdd(i, i);
}
System.Console.WriteLine("ThreadWork1 run } ");
}
}
class ThreadWork2//consumer
{
public ThreadWork2()
{ }
public void run()
{
bool IsOk = false;
Console.WriteLine("ThreadWork2 run { ");
int nCnt = 0,i=0,nValue=http://www.mamicode.com/0;
while (nCnt<100)
{
IsOk = _TestDictionary.TryGetValue(i,out nValue);
if (IsOk)
{
Console.WriteLine("ThreadWork2 consumer: "+i*i+"====="+i);
nValue *= nValue;
nCnt++;
i++;
}
}
Console.WriteLine("ThreadWork2 run } ");
}
}
static void StartT1()
{
ThreadWork1 work1 = new ThreadWork1();
work1.run();
}
static void StartT2()
{
ThreadWork2 work2 = new ThreadWork2();
work2.run();
}
static void Main(string[] args)
{
Task t1 = new Task(() => StartT1());
Task t2 = new Task(() => StartT2());
bool bIsNext = true;
int nValue = 0;
_TestDictionary = new ConcurrentDictionary<int, int>();
Console.WriteLine("Sample 4-5 Main {");
Console.WriteLine("Main t1 t2 started {");
t1.Start();
t2.Start();
Console.WriteLine("Main t1 t2 started }");
Console.WriteLine("Main wait t1 t2 end {");
Task.WaitAll(t1, t2);
Console.WriteLine("Main wait t1 t2 end }");
foreach (var pair in _TestDictionary)
{
Console.WriteLine(pair.Key + " : " + pair.Value);
}
IEnumerator<KeyValuePair<int, int>> enumer = _TestDictionary.GetEnumerator();
while (bIsNext)
{
bIsNext = enumer.MoveNext();
Console.WriteLine("Key: " + enumer.Current.Key +" Value: " + enumer.Current.Value);
_TestDictionary.TryRemove(enumer.Current.Key, out nValue);
}
Console.WriteLine("\n\nDictionary Count: " + _TestDictionary.Count);
Console.WriteLine("Sample 4-5 Main }");
Console.ReadKey();
}
}
}
总结说明: .NET4包括的新命名空间System.Collection,Concurrent有几个线程安全的集合类.线程安全的集合可防止多个线程以相互冲突的方式訪问集合.
为了对集合进行线程安全的訪问,定义了IPriducerConsumerCollection<T>接口.这个接口中最重要的方法是TryAdd()和TryTake().TryAdd()方法尝试给集合加入一项,但假设集合禁止加入项,这个操作可能失败.为了给出相关信息,TryAdd()方法返回一个布尔值,以说明操作是成功还是失败.TryTake()方法也以这样的方式工作,以通过调用者操作是成功还是失败,并在操作成功时返回集合中的项.
C#高级编程五十八天----并行集合