Plinq和linq性能比较c＃

Question

有人可以为我解释为什么PLinq在第一个示例中比正常的linq更好，而在第二个示例中最差？我看到的唯一区别是ExpensiveComputation（）函数中的Thread.Sleep（）

using System;
using System.Collections.Generic;
using System.Diagnostics;
using System.Linq;
using System.Threading;

namespace PlinqTest
{
    public class Employee
    {
        public string Name { get; set; }
        public double Salary { get; set; }
        public bool ExpensiveComputation()
        {
            Thread.Sleep(10);
            return (Salary > 2000 && Salary < 3000);
        }
        public bool NonExpensiveComputation()
        {
            return (Salary > 2000 && Salary < 3000);
        }
    }

    class Program
    {
        static void Main(string[] args)
        {
            List<Employee> employeeList = GetData();
            Stopwatch stopWatch = new Stopwatch();

            // Example 1

            stopWatch.Start();
            var linqResult = employeeList.Where<Employee>(e => e.ExpensiveComputation());
            int empCount = linqResult.Count();
            stopWatch.Stop();
            Console.WriteLine(string.Format("Time taken by old LINQ in Expensive Computation is {0} to get {1} Employees", stopWatch.Elapsed.TotalMilliseconds, empCount));
            stopWatch.Reset();
            stopWatch.Start();
            linqResult = employeeList.AsParallel<Employee>().Where<Employee>(e => e.ExpensiveComputation());
            empCount = linqResult.Count();
            stopWatch.Stop();
            Console.WriteLine(string.Format("Time taken by new PLINQ in Expensive Computation is {0} to get {1} Employees", stopWatch.Elapsed.TotalMilliseconds, empCount));
            stopWatch.Reset();
            Console.WriteLine();

            // Example 2

            stopWatch.Start();
            linqResult = employeeList.Where<Employee>(e => e.NonExpensiveComputation());
            empCount = linqResult.Count();
            stopWatch.Stop();
            Console.WriteLine(string.Format("Time taken by old LINQ in Non Expensive Computation is {0} to get {1} Employees", stopWatch.Elapsed.TotalMilliseconds, empCount));
            stopWatch.Reset();
            stopWatch.Start();
            linqResult = employeeList.AsParallel<Employee>().Where<Employee>(e => e.NonExpensiveComputation());
            empCount = linqResult.Count();
            stopWatch.Stop();
            Console.WriteLine(string.Format("Time taken by new PLINQ in Non Expensive Computation is {0} to get {1} Employees", stopWatch.Elapsed.TotalMilliseconds, empCount));
            stopWatch.Reset();
            Console.ReadKey();
        }

        static List<Employee> GetData()
        {
            List<Employee> employeeList = new List<Employee>();
            Random random = new Random(1000);
            for (int i = 0; i < 500; i++)
            {
                employeeList.Add(new Employee() { Name = "Employee" + i, Salary = GetRandomNumber(random, 1000, 5000) });
            }
            return employeeList;
        }
        static double GetRandomNumber(Random random, double minimum, double maximum)
        {
            return random.NextDouble() * (maximum - minimum) + minimum;
        }
    }
}

这里是结果：

旧的LINQ在昂贵的计算中花费的时间为5458.0269以获取135名员工

新的PLINQ在昂贵的计算中花费的时间为741.835，以获得135名员工

旧LINQ在非昂贵计算中花费的时间为0.3653以获取135名员工

新的PLINQ在非昂贵的计算中花费的时间为0.9175，以获得135名员工

Answer 1

平行主义不是免费的，它有间接费用。如果您所做的工作量很小，那么间接费用可能会比节省的时间更多。

我强烈建议您阅读免费的电子书Patterns of Parallel Programming，它讨论了您遇到的问题，并很好地解释了它们。