与std :: unordered_map或std :: map相比,MFC CMap是否具有良好的性能

Question

不MFC CMap相比有不错的表现std::unordered_map还是std::map,我问这个问题,因为我打算在我公司启动一个项目,并加快我要开始发展与现有的"类似"的项目,但在最后,有MFC CMap(哈希表映射)ans我认为使用std::unordered_map可以增加性能.我没有找到任何与CMap互联网相关的基准或好文章.否则,std::unordered_map我是否必须修改哈希表的大小,CMap以避免冲突和性能问题？

Answer 1

我做了非常简单的性能对比测试:

int nElements = 1000000;
CMap<int, int, CString, LPCTSTR> MfcHashTable;
MfcHashTable.InitHashTable(nElements);

// CMap insert
DWORD dwStart = ::GetTickCount();
for(int i=0; i<nElements; i++)
{
    CString sBase;
    sBase.AppendFormat(_T("Test String %d"), i);
    MfcHashTable[i] = sBase;
}

DWORD dwMfcMapInsert = ::GetTickCount() - dwStart;

// CMap lookup
CString sValue;

dwStart = ::GetTickCount();
for(int i=0; i<nElements; i++)
{
    MfcHashTable.Lookup(i, sValue);
}
DWORD dwMfcMapLookup = ::GetTickCount() - dwStart;

// std::map insert
std::map<int, CString> StdMap;
dwStart = ::GetTickCount();
for(int i=0; i<nElements; i++)
{
    CString sBase;
    sBase.AppendFormat(_T("Test String %d"), i);
    StdMap[i] = sBase;
}
DWORD dwStdMapInsert = ::GetTickCount() - dwStart;

//std::map lookup
dwStart = ::GetTickCount();
std::map<int, CString>::iterator it;
for(int i=0; i<nElements; i++)
{
    it = StdMap.find(i);
    CString sBase = it->second;
}
DWORD dwStdMapLookup = ::GetTickCount() - dwStart;

// std::unordered_map insert (hash table)
std::unordered_map<int, CString> StdUnordMap;
dwStart = ::GetTickCount();
for(int i=0; i<nElements; i++)
{
    CString sBase;
    sBase.AppendFormat(_T("Test String %d"), i);
    StdUnordMap[i] = sBase;
}
DWORD dwStdUnordMapInsert = ::GetTickCount() - dwStart;

//std::map lookup
dwStart = ::GetTickCount();
std::unordered_map<int, CString>::iterator it1;
for(int i=0; i<nElements; i++)
{
    it1 = StdUnordMap.find(i);
    CString sBase = it1->second;
}
DWORD dwStdUnordMapLookup = ::GetTickCount() - dwStart;

cout << dwMfcMapInsert << endl;
cout << dwMfcMapLookup << endl;

cout << dwStdMapInsert << endl;
cout << dwStdMapLookup << endl;

cout << dwStdUnordMapInsert << endl;
cout << dwStdUnordMapLookup << endl;

以下是英特尔酷睿i5 2.5Ghz 8GB内存(联想ThinkPad X230)上1000000个元素的结果:

MFC CMap insert: 1125
MFC CMap lookup: 125
std::map insert: 1406
std::map lookup: 172
std::unordered_map insert: 1578
std::unordered_map lookup: 140

令人惊讶的CMap是,这里是胜利者.事实证明,丑陋的遗产CMap毕竟不是那么糟糕!