为什么并行代码更慢? [英] Why parallel code is slower?

问题描述

我用大数组创建了一个简单的测试，我得到了并行 for 和普通 for 之间的巨大差异

I created a simple test with big arrays, and I get a big difference between parallel for and normal for

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

namespace ConsoleApplication1
{
    class Program
    {
        static void Main(string[] args)
        {
            var rn = new Random(541);

            var tmStart = DateTime.Now;
            int[,] a = new int[2048, 2048];
            for (int i = 0; i < 2048; i++)
                for (int j = 0; j < 2048; j++)
                    a[i, j] = rn.Next();
            var tmEnd = DateTime.Now - tmStart;
            Console.WriteLine("Normal..: {0}",tmEnd.ToString());

            tmStart = DateTime.Now;
            a = new int[2048, 2048];
            Parallel.For(0, 2048, i =>
                {
                    for (int j = 0; j < 2048; j++)
                        a[i, j] = rn.Next();
                });
            tmEnd = DateTime.Now - tmStart;
            Console.WriteLine("Parallel: {0}", tmEnd.ToString());
        }
    }
}

处理时间:

正常..:00:00:00.1250071
并行:00:00:00.3880222

Normal..: 00:00:00.1250071
Parallel: 00:00:00.3880222

为什么差别这么大?

我想象如果你使用多个线程，它会更快......

I imagined that if you use several threads, it would be faster...

推荐答案

您的基准测试存在一些问题:

There are a few things wrong with your benchmark:

1. 您的任务非常小，线程开销很大.
2. 您使用 Random 的实例打破了线程规则.谁知道这有什么影响?
3. 你的时机很粗糙.

1. Your tasks are so small that the threading overhead is significant.
2. You break the threading rules with the instance of Random. Who knows what impact this has?
3. Your timing is, well, crude.

这是一个更好的基准测试，显示代码的并行版本速度更快:

Here's a better benchmark that shows the parallel version of the code being faster:

using System;
using System.Diagnostics;
using System.Threading.Tasks;

namespace ConsoleApplication1
{
    class Program
    {
        const int N = 1024;

        static void Main(string[] args)
        {
            int[,] a = new int[N, N];

            var stopwatch = Stopwatch.StartNew();
            for (int i = 0; i < N; i++)
                for (int k = 0; k < N; k++)
                    for (int j = 0; j < N; j++)
                        a[i, j] = i+j;
            Console.WriteLine("Normal..: {0}", stopwatch.ElapsedMilliseconds);

            stopwatch = Stopwatch.StartNew();
            Parallel.For(0, N, i =>
            {
                for (int k = 0; k < N; k++)
                    for (int j = 0; j < N; j++)
                        a[i, j] = i+j;
            });
            Console.WriteLine("Parallel: {0}", stopwatch.ElapsedMilliseconds);

            Console.ReadLine();
        }
    }
}

这是使用随机数生成器的一个:

And here's one which uses random number generators:

using System;
using System.Diagnostics;
using System.Threading.Tasks;

namespace ConsoleApplication1
{
    class Program
    {
        const int N = 512;

        static void Main(string[] args)
        {
            int[,] a = new int[N, N];

            var stopwatch = Stopwatch.StartNew();
            var rng = new Random();
            for (int i = 0; i < N; i++)
                for (int k = 0; k < N; k++)
                    for (int j = 0; j < N; j++)
                        a[i, j] = rng.Next();
            Console.WriteLine("Normal..: {0}", stopwatch.ElapsedMilliseconds);

            stopwatch = Stopwatch.StartNew();
            Parallel.For(0, N, i =>
            {
                var rngpar = new Random();
                for (int k = 0; k < N; k++)
                    for (int j = 0; j < N; j++)
                        a[i, j] = rngpar.Next();
            });
            Console.WriteLine("Parallel: {0}", stopwatch.ElapsedMilliseconds);

            Console.ReadLine();
        }
    }
}

这篇关于为什么并行代码更慢?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！