none
关于并行Linq,手动分区. RRS feed

 > 

  • 问题

  • Question
    登录进行投票
    0
    登录进行投票 
    static void Main(string[] args) {
            int arraySize = 100000000;
            int[] intArray = new int[arraySize];
            Random r = new Random();    //随机给数组赋值.
            for(int i = 0; i < intArray.Length; i++) {
                intArray[i] = r.Next(100);
            }
            DateTime start = DateTime.Now;  //开始计时.
            var query = (from x in intArray //方式(一):普通模式.
                         where x < 50
                         select x).Sum();
            DateTime part = DateTime.Now;    //结束计时.
            Console.WriteLine("耗时: " + (part - start).TotalSeconds);
            //耗时: 2.203126.
            //自动分区:
            var query2 = (from x in intArray.AsParallel()   //方式(二):默认情况下的平行Linq.
                         where x < 50
                         select x).Sum();
            DateTime end = DateTime.Now;    //结束计时.
            Console.WriteLine("耗时: " + (end - part).TotalSeconds);
            //耗时: 1.503086.
            //手动分区:
            var queryManual = (from x in Partitioner.Create(intArray, true).AsParallel()    //手动分区的平行Linq.
                               where x < 50
                               select x).Sum();
            DateTime end2 = DateTime.Now;    //结束计时.
            Console.WriteLine("耗时: " + (end2 - end).TotalSeconds);
            //耗时: 2.2711299.
            Console.ReadKey();
        }
    //大家懂的...
这三种方式...
我郁闷的是,手动分区的耗时和普通模式耗时一样的...
是不是这不是分区哈...如果是分区...都没有用的..赤裸裸的多余...



    2013年6月15日 6:17
    |

答案

  • Question
    登录进行投票
    0
    登录进行投票

    一般只需要使用AsParallel进行自动产生线程自动计算，无须使用Partition分区机制。

    而分区往往和Parallel.ForEach使用，对每一块自身数据都是很大的情况下很适合：

    namespace Csharp
{
    public class HeaveyWork
    {
        public static Random rd = new Random(DateTime.Now.GetHashCode());
 
        public void Wait()
        {
            //长时间工作
            Thread.Sleep(rd.Next(10,100));
        }
    }
    public class Program
    {
        static void Main(string[] args)
        {
            HeaveyWork[]works = new HeaveyWork[100];
            for (int i = 0; i < works.Length; i++)
            {
                works[i] = new HeaveyWork();
            }
            Stopwatch watch = new Stopwatch();
            watch.Start();
            Parallel.ForEach<HeaveyWork>(Partitioner.Create(works, true), (work) => work.Wait());
            watch.Stop();
            Console.WriteLine(watch.ElapsedMilliseconds);
        }
    }
}

    If you think one reply solves your problem, please mark it as An Answer, if you think someone's reply helps you, please mark it as a Proposed Answer

    Help by clicking：
    Click here to donate your rice to the poor
    Click to Donate
    Click to feed Dogs & Cats

    Found any spamming-senders? Please report at: Spam Report

    2013年6月15日 10:11
    |
    版主