并行编程_Parallel

一、Parallel.Invoke执行多个方法

没有特定执行顺序，利用cpu多核并发执行,在运行并行方法前都会产生一些额外的开销,我们一定要测量运行结果相比没有并行是否有优化

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

#region Parallel.Invoke           Stopwatch watch = new Stopwatch();           watch.Start();           Parallel.Invoke(Pinvoke1, Pinvoke2);           watch.Stop();           Console.WriteLine("Parallel run " + watch.ElapsedMilliseconds + " ms");           watch.Reset();           watch.Start();           Pinvoke1();           Pinvoke2();           watch.Stop();           Console.WriteLine("Normal run " + watch.ElapsedMilliseconds + " ms");           Console.ReadKey();             #endregion Parallel.Invoke

?

1 2 3 4 5 6 7 8 9 10 11

private static void Pinvoke1()       {           Thread.Sleep(3000);           Console.WriteLine("Pinvoke1 ThreadId="+Thread.CurrentContext.ContextID);       }         private static void Pinvoke2()       {           Thread.Sleep(2000);           Console.WriteLine("Pinvoke2 ThreadId=" + Thread.CurrentContext.ContextID);       }

　结果：

二、Parallel.For   每一次迭代一会当作一个任务交给一个线程，线程可能会重用，cpu决定

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

Stopwatch watch = new Stopwatch();          watch.Start();            for (int i = 0; i < 5; i++)          {              Console.WriteLine(i);              Thread.Sleep(1000);          }          watch.Stop();          Console.WriteLine(watch.Elapsed);                watch.Restart();          Parallel.For(0, 5, i =>          {              Console.WriteLine("===" + Thread.CurrentThread.ManagedThreadId + "===");              Console.WriteLine(i);              Thread.Sleep(1000);          });          watch.Stop();          Console.WriteLine(watch.Elapsed);          Console.ReadKey();

效果还是非常明显的,但是有个问题是Parallel.For是并行执行,每次迭代基本都会启用新线程,那么我们共用一个全局变量的时候,必须要加锁,出现资源等待,这时用并行就不合适了:示例如下:

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27

object asyncObj = new object();             long value = 0;             Stopwatch watch = new Stopwatch();             watch.Start();             for (int i = 0; i < 9999; i++)             {                 for (int j = 0; j < 19999; j++)                 {                     value++;                 }             }             watch.Stop();             Console.WriteLine("正常运行是" + watch.ElapsedMilliseconds + " ms");             watch.Restart();             Parallel.For(0, 9999, i =>             {                 for (int j = 0; j < 19999; j++)                 {                     lock (asyncObj)                     {                         value++;                     }                 }             });             watch.Stop();             Console.WriteLine("并行运行是" + watch.ElapsedMilliseconds + " ms");             Console.ReadKey();

三、Parallel.ForEach  都foreach一样，只不过是并行的，也是每次循环一个线程处理

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

List<Person> list = new List<Person>();         for (int i = 1; i < 100; i++)         {             list.Add(new Person             {                 Age = i,                 Name = "Name" + i             });         }         Stopwatch watch = new Stopwatch();         watch.Start();         Parallel.ForEach(list, p =>         {             Console.WriteLine("===" + Thread.CurrentThread.ManagedThreadId + "===");             Console.WriteLine(p.Name);         });         watch.Stop();         Console.ReadKey();

四、并行中断

ParallelLoopState:

1、Break-并行循环执行了当前迭代后尽快地停止执行。

//更改方法stop下面的示例,运行结果不一定哦,49,50都出现了

2、Stop-直接停止

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14

//结果是50   List<Person> list=new List<Person>();             Parallel.For(0, 100, (num, state) =>             {                 if (num<50)                 {                     list.Add(new Person() {Age = num});                 }                 else                 {                     state.Stop();                 }             });             Console.WriteLine(list.Count);

五、异常处理

并行执行的时候使用AggregateException获取全部异常,用Exception只能获取其中一个;

?

1 2 3 4 5 6 7 8 9 10 11

try {     Parallel.Invoke(Pinvoke1, Pinvoke2); } catch (AggregateException aggregateException) {     foreach (var innerException in aggregateException.InnerExceptions)     {         Console.WriteLine(innerException.Message);     } }

?

1 2 3 4 5 6 7 8 9 10 11 12 13

private static void Pinvoke1()     {         Thread.Sleep(3000);         Console.WriteLine("Pinvoke1 ThreadId=" + Thread.CurrentContext.ContextID);         throw new Exception("并行任务1抛出异常");     }       private static void Pinvoke2()     {         Thread.Sleep(2000);         Console.WriteLine("Pinvoke2 ThreadId=" + Thread.CurrentContext.ContextID);         throw new Exception("并行任务2抛出异常");     }

 六、并行Linq （AsParallel，可以应用任何集合，提高处理速度）

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Stopwatch watch=new Stopwatch();           List<Person> list = new List<Person>();           Parallel.For(0,10000,p =>           {               list.Add(new Person() {Age = p,Name = "name"+p});//注意并行会出现有空对象的现象;也可以调用Invoke方法,里面写for循环就不会出现空对象           });           Console.WriteLine(list.Count);           watch.Start();           Console.WriteLine(list.Count);           var persons = list.Where(c=>c!=null&&c.Age>100).ToList();           watch.Stop();           Console.WriteLine(watch.ElapsedMilliseconds);           watch.Restart();           var parallelPersons = list.AsParallel().Where(c=> c != null && c.Age>100).ToList();           watch.Stop();           Console.WriteLine(watch.ElapsedMilliseconds);           Console.ReadKey();