使用GetLogicalProcessorInformation获取逻辑处理器的详细信息（NUMA节点数、物理CPU数、CPU核心数、逻辑CPU数、各级Cache）

　　现在多核处理器已经很普及了，市场主流是双核处理器，还有4核、8核等高端产品。而且Intel推广了超线程技术（Hyper-Threading Technology, HTT），可以将一个物理核心模拟为两个逻辑处理器。这一切使得“CPU数量”这一概念变得复杂起来，对于软件开发人员来说，希望能获得物理CPU数、CPU核心数、逻辑CPU数等详细信息。
　　在Windows平台，可以调用GetLogicalProcessorInformation函数来获取它们的详细信息。

一、背景知识

　　先来明确一下名词——
physical processor packages：物理处理器封装个数，即俗称的“物理CPU数”。例如一块“Intel Core i3-2310M”只有1个“物理处理器封装个数”。若对于有多个处理器插槽的服务器，“物理处理器封装个数”很可能会大于1。
processor cores：处理器核心数，即俗称的“CPU核心数”。例如“Intel Core i3-2310M”是双核处理器，它有2个“处理器核心数”。
logical processors：逻辑处理器数，即俗称的“逻辑CPU数”。例如“Intel Core i3-2310M”支持超线程，一个物理核心能模拟为两个逻辑处理器，即一块“Intel Core i3-2310M”有4个“逻辑处理器数”。

　　再来看看2个大家可能不太熟悉的名词——
SMP：Symmetrical Multi-Processing，对称多处理机。
NUMA：Non Uniform Memory Access，非均匀访存模型。http://msdn.microsoft.com/en-us/library/aa363804(v=vs.85).aspx

　　这个两个名词牵涉到很多专业知识，这里不做详细介绍，感兴趣的同学可以自行翻阅相关资料。
　　老版本的Windows系统（例如Windows XP）采用的是SMP模型。但后来因多核处理器及异构计算的发展，从Windows Server 2003开始使用NUMA模型，系统中支持多个NUMA节点。对于开发人员来说，当只有1个NUMA节点时，与SMP模型是差不多的。
　　对于 Windows XP，在打上SP3补丁后，也可以利用GetLogicalProcessorInformation函数获得NUMA等信息。

二、GetLogicalProcessorInformation函数的使用心得

　　在MSDN上我们可以查到GetLogicalProcessorInformation函数的帮助——
http://msdn.microsoft.com/en-us/library/ms683194(v=vs.85).aspx
GetLogicalProcessorInformation function

　　GetLogicalProcessorInformation函数还牵涉到一些结构体和枚举——
http://msdn.microsoft.com/en-us/library/ms686694(v=vs.85).aspx
SYSTEM_LOGICAL_PROCESSOR_INFORMATION structure

http://msdn.microsoft.com/en-us/library/ms684197(v=vs.85).aspx
LOGICAL_PROCESSOR_RELATIONSHIP enumeration

http://msdn.microsoft.com/en-us/library/ms681979(v=vs.85).aspx
CACHE_DESCRIPTOR structure

http://msdn.microsoft.com/en-us/library/ms684844(v=vs.85).aspx
PROCESSOR_CACHE_TYPE enumeration

　　GetLogicalProcessorInformation函数用起来是有一定复杂性的。因为它返回的是SYSTEM_LOGICAL_PROCESSOR_INFORMATION数组，数组中的每一项分别描述了不同的信息，学习曲线较陡峭。
　　虽然MSDN上有该函数的范例代码，但是它屏蔽了很多细节，对我们的帮助有限。于是我将该范例程序作了改进，显示了SYSTEM_LOGICAL_PROCESSOR_INFORMATION数组中每一项的详细信息。

　　心得——
1.SYSTEM_LOGICAL_PROCESSOR_INFORMATION结构ProcessorMask是ULONG_PTR类型的。在32位系统上是32位，64位系统上是64位。为了简化代码，建议强制转型为UINT64类型，调用printf等输出函数时使用“I64”格式码。
2.ProcessorMask是处理器掩码，每一位代表一个逻辑处理器。所以一般来说，32位系统最多支持32个逻辑处理器，64位系统最多支持64个逻辑处理器。
3.对于Windows 7和Windows Server 2008 R2来说，能突破64个逻辑处理器限制，最高支持256个逻辑处理器。新加了 处理器组（Processor Groups）概念，详见：http://msdn.microsoft.com/en-us/library/dd405503(v=vs.85).aspx。

三、全部代码

　　全部代码——

#include <windows.h>
#include <malloc.h>    
#include <stdio.h>
#include <tchar.h>

#if (_WIN32_WINNT < 0x0600)	// [zyl910] 低版本的Windows SDK没有定义 RelationProcessorPackage 等常量
#define RelationProcessorPackage	3
#define RelationGroup	4
#endif
// [zyl910] LOGICAL_PROCESSOR_RELATIONSHIP枚举的名称
const LPTSTR Names_LOGICAL_PROCESSOR_RELATIONSHIP[] = {
	_T("RelationProcessorCore")
	,_T("RelationNumaNode")
	,_T("RelationCache")
	,_T("RelationProcessorPackage")
	,_T("RelationGroup")
};
// [zyl910] PROCESSOR_CACHE_TYPE枚举的名称
const LPTSTR Names_PROCESSOR_CACHE_TYPE[] = {
	_T("CacheUnified")
	,_T("CacheInstruction")
	,_T("CacheData")
	,_T("CacheTrace")
};

typedef BOOL (WINAPI *LPFN_GLPI)(
    PSYSTEM_LOGICAL_PROCESSOR_INFORMATION, 
    PDWORD);


// Helper function to count set bits in the processor mask.
DWORD CountSetBits(ULONG_PTR bitMask)
{
    DWORD LSHIFT = sizeof(ULONG_PTR)*8 - 1;
    DWORD bitSetCount = 0;
    ULONG_PTR bitTest = (ULONG_PTR)1 << LSHIFT;    
    DWORD i;
    
    for (i = 0; i <= LSHIFT; ++i)
    {
        bitSetCount += ((bitMask & bitTest)?1:0);
        bitTest/=2;
    }

    return bitSetCount;
}

int _cdecl _tmain ()
{
    LPFN_GLPI glpi;
    BOOL done = FALSE;
    PSYSTEM_LOGICAL_PROCESSOR_INFORMATION buffer = NULL;
    PSYSTEM_LOGICAL_PROCESSOR_INFORMATION ptr = NULL;
    DWORD returnLength = 0;
    DWORD logicalProcessorCount = 0;
    DWORD numaNodeCount = 0;
    DWORD processorCoreCount = 0;
    DWORD processorL1CacheCount = 0;
    DWORD processorL2CacheCount = 0;
    DWORD processorL3CacheCount = 0;
    DWORD processorPackageCount = 0;
    DWORD byteOffset = 0;
    PCACHE_DESCRIPTOR Cache;

    glpi = (LPFN_GLPI) GetProcAddress(
                            GetModuleHandle(TEXT("kernel32")),
                            "GetLogicalProcessorInformation");
    if (NULL == glpi) 
    {
        _tprintf(TEXT("\nGetLogicalProcessorInformation is not supported.\n"));
        return (1);
    }

    while (!done)
    {
        DWORD rc = glpi(buffer, &returnLength);

        if (FALSE == rc) 
        {
            if (GetLastError() == ERROR_INSUFFICIENT_BUFFER) 
            {
                if (buffer) 
                    free(buffer);

                buffer = (PSYSTEM_LOGICAL_PROCESSOR_INFORMATION)malloc(
                        returnLength);

                if (NULL == buffer) 
                {
                    _tprintf(TEXT("\nError: Allocation failure\n"));
                    return (2);
                }
            } 
            else 
            {
                _tprintf(TEXT("\nError %d\n"), GetLastError());
                return (3);
            }
        } 
        else
        {
            done = TRUE;
        }
    }

    ptr = buffer;

	if (true)	// [zyl910] 显示SYSTEM_LOGICAL_PROCESSOR_INFORMATION结构体的详细信息
	{
		DWORD cnt = returnLength / sizeof(SYSTEM_LOGICAL_PROCESSOR_INFORMATION);	// 计算SYSTEM_LOGICAL_PROCESSOR_INFORMATION结构体的数目
		for(DWORD i=0; i<cnt; ++i)
		{
			_tprintf(TEXT("SYSTEM_LOGICAL_PROCESSOR_INFORMATION[%d]\n"), i);
			_tprintf(TEXT("\t.ProcessorMask:\t0x%.16I64X\t//%I64d\n"), (UINT64)ptr[i].ProcessorMask, (UINT64)ptr[i].ProcessorMask);
			_tprintf(TEXT("\t.Relationship:\t%d\t//%s\n"), ptr[i].Relationship, Names_LOGICAL_PROCESSOR_RELATIONSHIP[max(0,min(ptr[i].Relationship, RelationGroup))]);
			for(int j=0; j<2; ++j)	_tprintf(TEXT("\t.Reserved[%d]:\t//0x%.16I64X\t%I64d\n"), j, (UINT64)ptr[i].Reserved[j], (UINT64)ptr[i].Reserved[j]);
			if (RelationCache==ptr[i].Relationship)
			{
				_tprintf(TEXT("\t.Cache.Level:\t%u\n"), ptr[i].Cache.Level);
				_tprintf(TEXT("\t.Cache.Associativity:\t0x%.2X\t//%u\n"), ptr[i].Cache.Associativity, ptr[i].Cache.Associativity);
				_tprintf(TEXT("\t.Cache.LineSize:\t0x%.4X\t//%u\n"), ptr[i].Cache.LineSize, ptr[i].Cache.LineSize);
				_tprintf(TEXT("\t.Cache.Size:\t0x%.8X\t//%u\n"), ptr[i].Cache.Size, ptr[i].Cache.Size);
				_tprintf(TEXT("\t.Cache.Type:\t%d\t//%s\n"), ptr[i].Cache.Type, Names_PROCESSOR_CACHE_TYPE[max(0,min(ptr[i].Cache.Type, CacheTrace))]);
			}
		}
	}

    while (byteOffset + sizeof(SYSTEM_LOGICAL_PROCESSOR_INFORMATION) <= returnLength) 
    {
        switch (ptr->Relationship) 
        {
        case RelationNumaNode:
            // Non-NUMA systems report a single record of this type.
            numaNodeCount++;
            break;

        case RelationProcessorCore:
            processorCoreCount++;

            // A hyperthreaded core supplies more than one logical processor.
            logicalProcessorCount += CountSetBits(ptr->ProcessorMask);
            break;

        case RelationCache:
            // Cache data is in ptr->Cache, one CACHE_DESCRIPTOR structure for each cache. 
            Cache = &ptr->Cache;
            if (Cache->Level == 1)
            {
                processorL1CacheCount++;
            }
            else if (Cache->Level == 2)
            {
                processorL2CacheCount++;
            }
            else if (Cache->Level == 3)
            {
                processorL3CacheCount++;
            }
            break;

        case RelationProcessorPackage:
            // Logical processors share a physical package.
            processorPackageCount++;
            break;

        default:
            _tprintf(TEXT("\nError: Unsupported LOGICAL_PROCESSOR_RELATIONSHIP value.\n"));
            break;
        }
        byteOffset += sizeof(SYSTEM_LOGICAL_PROCESSOR_INFORMATION);
        ptr++;
    }

    _tprintf(TEXT("\nGetLogicalProcessorInformation results:\n"));
    _tprintf(TEXT("Number of NUMA nodes: %d\n"), 
             numaNodeCount);
    _tprintf(TEXT("Number of physical processor packages: %d\n"), 
             processorPackageCount);
    _tprintf(TEXT("Number of processor cores: %d\n"), 
             processorCoreCount);
    _tprintf(TEXT("Number of logical processors: %d\n"), 
             logicalProcessorCount);
    _tprintf(TEXT("Number of processor L1/L2/L3 caches: %d/%d/%d\n"), 
             processorL1CacheCount,
             processorL2CacheCount,
             processorL3CacheCount);
    
    free(buffer);

    return 0;
}

 

四、输出信息

　　例如我的处理器是“Intel Core i3-2310M”，该程序的输出信息为——

SYSTEM_LOGICAL_PROCESSOR_INFORMATION[0]
	.ProcessorMask:	0x0000000000000005	//5
	.Relationship:	0	//RelationProcessorCore
	.Reserved[0]:	//0x0000000000000001	1
	.Reserved[1]:	//0x0000000000000000	0
SYSTEM_LOGICAL_PROCESSOR_INFORMATION[1]
	.ProcessorMask:	0x0000000000000005	//5
	.Relationship:	2	//RelationCache
	.Reserved[0]:	//0x0000800000400801	140737492551681
	.Reserved[1]:	//0x0000000000000002	2
	.Cache.Level:	1
	.Cache.Associativity:	0x08	//8
	.Cache.LineSize:	0x0040	//64
	.Cache.Size:	0x00008000	//32768
	.Cache.Type:	2	//CacheData
SYSTEM_LOGICAL_PROCESSOR_INFORMATION[2]
	.ProcessorMask:	0x0000000000000005	//5
	.Relationship:	2	//RelationCache
	.Reserved[0]:	//0x0000800000400801	140737492551681
	.Reserved[1]:	//0x0000000000000001	1
	.Cache.Level:	1
	.Cache.Associativity:	0x08	//8
	.Cache.LineSize:	0x0040	//64
	.Cache.Size:	0x00008000	//32768
	.Cache.Type:	1	//CacheInstruction
SYSTEM_LOGICAL_PROCESSOR_INFORMATION[3]
	.ProcessorMask:	0x0000000000000005	//5
	.Relationship:	2	//RelationCache
	.Reserved[0]:	//0x0004000000400802	1125899911038978
	.Reserved[1]:	//0x0000000000000000	0
	.Cache.Level:	2
	.Cache.Associativity:	0x08	//8
	.Cache.LineSize:	0x0040	//64
	.Cache.Size:	0x00040000	//262144
	.Cache.Type:	0	//CacheUnified
SYSTEM_LOGICAL_PROCESSOR_INFORMATION[4]
	.ProcessorMask:	0x000000000000000F	//15
	.Relationship:	3	//RelationProcessorPackage
	.Reserved[0]:	//0x0000000000000000	0
	.Reserved[1]:	//0x0000000000000000	0
SYSTEM_LOGICAL_PROCESSOR_INFORMATION[5]
	.ProcessorMask:	0x000000000000000A	//10
	.Relationship:	0	//RelationProcessorCore
	.Reserved[0]:	//0x0000000000000001	1
	.Reserved[1]:	//0x0000000000000000	0
SYSTEM_LOGICAL_PROCESSOR_INFORMATION[6]
	.ProcessorMask:	0x000000000000000A	//10
	.Relationship:	2	//RelationCache
	.Reserved[0]:	//0x0000800000400801	140737492551681
	.Reserved[1]:	//0x0000000000000002	2
	.Cache.Level:	1
	.Cache.Associativity:	0x08	//8
	.Cache.LineSize:	0x0040	//64
	.Cache.Size:	0x00008000	//32768
	.Cache.Type:	2	//CacheData
SYSTEM_LOGICAL_PROCESSOR_INFORMATION[7]
	.ProcessorMask:	0x000000000000000A	//10
	.Relationship:	2	//RelationCache
	.Reserved[0]:	//0x0000800000400801	140737492551681
	.Reserved[1]:	//0x0000000000000001	1
	.Cache.Level:	1
	.Cache.Associativity:	0x08	//8
	.Cache.LineSize:	0x0040	//64
	.Cache.Size:	0x00008000	//32768
	.Cache.Type:	1	//CacheInstruction
SYSTEM_LOGICAL_PROCESSOR_INFORMATION[8]
	.ProcessorMask:	0x000000000000000A	//10
	.Relationship:	2	//RelationCache
	.Reserved[0]:	//0x0004000000400802	1125899911038978
	.Reserved[1]:	//0x0000000000000000	0
	.Cache.Level:	2
	.Cache.Associativity:	0x08	//8
	.Cache.LineSize:	0x0040	//64
	.Cache.Size:	0x00040000	//262144
	.Cache.Type:	0	//CacheUnified
SYSTEM_LOGICAL_PROCESSOR_INFORMATION[9]
	.ProcessorMask:	0x000000000000000F	//15
	.Relationship:	2	//RelationCache
	.Reserved[0]:	//0x0030000000400C03	13510798886308867
	.Reserved[1]:	//0x0000000000000000	0
	.Cache.Level:	3
	.Cache.Associativity:	0x0C	//12
	.Cache.LineSize:	0x0040	//64
	.Cache.Size:	0x00300000	//3145728
	.Cache.Type:	0	//CacheUnified
SYSTEM_LOGICAL_PROCESSOR_INFORMATION[10]
	.ProcessorMask:	0x000000000000000F	//15
	.Relationship:	1	//RelationNumaNode
	.Reserved[0]:	//0x0000000000000000	0
	.Reserved[1]:	//0x0000000000000000	0

GetLogicalProcessorInformation results:
Number of NUMA nodes: 1
Number of physical processor packages: 1
Number of processor cores: 2
Number of logical processors: 4
Number of processor L1/L2/L3 caches: 4/2/1

 

源码下载——
http://files.cnblogs.com/zyl910/showGetLogicalProcessorInformation.rar