C＃有没有比PLINQ快的东西，它可以平均处理1000万个项目数组

Question

我基本上是在尝试找到一种方法，以最快的速度进行分组时平均产生1000万个馆藏。 以下是我用作基准的代码，但我似乎找不到任何使它更快的方法。 我正在根据StopWatrch mainSW

这是我的Program.cs

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;

namespace SpeedTest
{
    class Program
    {
        static void Main(string[] args)
        {
            global.BuildItems();
            System.Diagnostics.Stopwatch mainSW = new System.Diagnostics.Stopwatch();
            mainSW.Start();
            baseLineTest.plinqGroup();
            Console.WriteLine("Press Return");
            Console.ForegroundColor = ConsoleColor.Green;
            Console.WriteLine("Main SW Elapsed:" + mainSW.Elapsed);
            Console.ReadLine();
        }
    }
}

这是我的global.cs

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;

namespace SpeedTest
{
    class global
    {
        //10 million
        public static long globalIteration = 10000000;

        private static string[] classOptions = { "AS", "CS", "LS", "PE", "WP", "LS" };
        public static Items[] items { get; set; }

        public static void BuildItems()
        {
            System.Diagnostics.Stopwatch sw = new System.Diagnostics.Stopwatch();

            Random r = new Random();
            Random r2 = new Random();

            Console.WriteLine("Building list");
            items = new Items[globalIteration];
            sw.Start();
            for (int i = 0; i < globalIteration; i++)
            {
                items[i] = new Items();
                items[i].cl = classOptions[r.Next(0, 5)];
                items[i].uc = Convert.ToDecimal(r.Next(300, 1000));
            }
            Console.WriteLine("Building list sw: " + sw.Elapsed);
        }

    }
    class Items
    {
        public decimal uc;
        public string cl;

    }
}

这是我的baseLineTest.cs

using System;
using System.Linq;
using System.Threading.Tasks;

namespace SpeedTest
{
    class baseLineTest
    {

        public static void plinqGroup()
        {

            System.Diagnostics.Stopwatch sw = new System.Diagnostics.Stopwatch();
            Console.WriteLine("This test is a plinq over an array");


            decimal avg = 0;
            decimal sum = 0;
            string clas = "";
            sw.Start();
            var list = global.items.AsParallel().GroupBy(d => d.cl)
    .Select(
        g => new
        {
            Key = g.Key,
            Value = g.Average(s => s.uc)
        });

            foreach (var item in list)
                Console.WriteLine(string.Format("{0} : {1} ", item.Key, item.Value));

            Console.WriteLine("PLinq Group elapsed : " + sw.Elapsed);


        }

    }


}

Answer 1

这取决于您对数据的了解程度。 如果使用数据的代码知道“ classOptions”，那么可以通过首先创建一个结果对象数组然后使用Parallel.ForEach来使其快两倍（至少在我的四核计算机上）。操作数组

    [TestMethod]
    public void MeasureParallelForEach()
    {
        string[] classOptions = { "AS", "CS", "LS", "PE", "WP", "LS" };
        global.BuildItems();
        System.Diagnostics.Stopwatch sw = new System.Diagnostics.Stopwatch();
        Console.WriteLine("This test is a plinq over an array");

        sw.Start();
        IDictionary<string, Group> groups = global.classOptions.Distinct().ToDictionary(x => x, x => new Group(x));
        Parallel.ForEach(global.items, d => groups[d.cl].Add(d.uc));
        foreach (var item in groups)
            Console.WriteLine(string.Format("{0} : {1} ", item.Key, item.Value.Average));
        Console.WriteLine("Parallel.ForEach elapsed : " + sw.Elapsed);
    }

    public class Group
    {
        private int count;
        private decimal sum;
        public Group(string key)
        {
            Key = key;
        }

        public void Add(decimal d)
        {
            sum += d;
            count++;
        }

        public string Key { get; private set; }
        public decimal Average { get { return count==0?0m:sum/count; }}
    }