将线程绑定在某个具体的CPU逻辑内核上运行

据说在运算过程中，逻辑内核的切换会造成额外的开销，影响程序运行效率。最近开始尝试仿真优化（Sim-Opt）方面的研究，刚好需求中涉及这一内容。所以进行了一些尝试。
先上代码：

using System.Runtime.InteropServices;
using System.Diagnostics;
using System.Management;

namespace BandThreadToCPU
{
    
    class Program
    {
    
        //SetThreadAffinityMask: Set hThread run on logical processer(LP:) dwThreadAffinityMask
        [DllImport("kernel32.dll")]
        static extern UIntPtr SetThreadAffinityMask(IntPtr hThread, UIntPtr dwThreadAffinityMask);

        //Get the handler of current thread
        [DllImport("kernel32.dll")]
        static extern IntPtr GetCurrentThread();

        //The real function
        public static void ChangeValue(object lpIdx)
        {
    
            //Bind current thread to a specific logical processer
            ulong LpId = SetCpuID((int)lpIdx);
            SetThreadAffinityMask(GetCurrentThread(), new UIntPtr(LpId));

            // Let program run for a while
            Stopwatch stopwatch = new Stopwatch();
            stopwatch.Start();
            for (int i = 0; i < 1000000; i++)
            {
    
                for (int j = 0; j < 1000000; j++)
                {
    
                    int _data = j;
                }
            }
            stopwatch.Stop();
            Console.WriteLine("Running Time: " + stopwatch.ElapsedMilliseconds.ToString());
        }

        //Get CPU id. index:0 -> id:1, 1->2, 2->4, 3->8, 4->16, ...
        static ulong SetCpuID(int lpIdx)
        {
    
            ulong cpuLogicalProcessorId = 0;
            if (lpIdx < 0 || lpIdx >= System.Environment.ProcessorCount)
            {
    
                lpIdx = 0;
            }
            cpuLogicalProcessorId |= 1UL << lpIdx;
            return cpuLogicalProcessorId;
        }

        static void Main(string[] args)
        {
    
            // Specify the cpu logical processor index you want. CPU lp index start from 0
            int LOGICAL_PROCESSOR_IDX = 4;

            // Get the ManagementClass object
            ManagementClass mc = new ManagementClass(new ManagementPath("Win32_Processor"));
            // Get the properties in the class
            ManagementObjectCollection moc = mc.GetInstances();

            foreach (ManagementObject mo in moc)
            {
    
                PropertyDataCollection properties = mo.Properties;
                // Print CPU properties
                foreach (PropertyData property in properties)
                {
    
                    Console.WriteLine( property.Name + ": " + property.Value);
                }
            }

            Stopwatch stopwatch = new Stopwatch();
            Thread thread = new Thread(new ParameterizedThreadStart(ChangeValue));
            stopwatch.Start();
            thread.Start(LOGICAL_PROCESSOR_IDX);
            thread.Join();
            stopwatch.Stop();
            Console.WriteLine("Total running time: " + stopwatch.ElapsedMilliseconds.ToString());
            Console.ReadKey();
        }
    }
}

其中：
SetThreadAffinityMask 将线程和core进行绑定。
GetCurrentThread 获取当前线程的句柄。
ChangeValue 真正的业务运算逻辑放在这里，此处是Demo，所以里面只写了一个很耗时的循环。
SetCpuID 这里是比较有意思的，CPU逻辑处理器索引（index）是从0开始的，但是SetThreadAffinityMask绑定时需要用到的是逻辑处理器的ID，index和id呈2的指数幂关系。即：idx=0对应id=1（2的0次方）；idx=3对应id=8（2的3次方）。

运行效果如下，可以看到，04号CPU core已经跑满，但其他core并没有来帮忙。

本篇内容主要参考了jimmy-yang的文章：《C#线程绑定到指定cpu》代码的主要部分也来源于此。主要的修在在于代码的简化，删除了与主要目的无关的部分，并修改了一些变量或函数的命名，修改了注释。使重点内容更加突出，容易理解。