递归函数严重降低性能

我有一个函数，我调用了无限多次（或者直到满足条件为止）。这个递归函数的问题是，在更高的级别上，它是由一个工作线程调用的，该线程推送并弹出到deque。与此同时，我正在用主线程迭代一个文件，这比递归函数快得多，递归函数处理通过主循环迭代获得的数据。下面是一个严重减慢速度的代码示例：

bool CCapsule::DecodeElementList( ElementList &elementList, char* pszBuffer, unsigned int uiBufferLength, std::string strSrcAddress, std::string strDestAddress, std::vector<std::string> &vsContents, bool bIsInner )
{
    // Create an buffer to contain pszBuffer.
    RBuffer rBuffer;
    rBuffer.data = pszBuffer;
    // Create an RElement List and Element Entry to store elements
    RRet rRet;
    RElementEntry rElementEntry;
    // Decode Element List
    if(rDecodeElementList(&m_rDecodeIter, &rElementList, 0) >= RET_SUCCESS)
    {
        std::vector<std::string> _vsContents;
        while ((RRet = rDecodeElementEntry(&m_rDecodeIter, &rElementEntry)) != RRET_END_OF_CONTAINER)
        {
            if (RRet < RET_SUCCESS)
            {
                return false;
            }
            std::string strEntryName = "";
            if (rElementEntry.data != NULL)
                rElementEntry.data;
            switch(rElementEntry.dataType)
            {
            case R_MSG:
                // Create a new RCapsule to encapsulate the inner message.
                m_pInnerMessage = new CCapsule();
                if ( false == m_pInnerMessage->Load( pszBuffer, uiBufferLength, strSrcAddress, strDestAddress, _vsContents, true ) )
                {
                    // An error occurred, clean up.
                    delete m_pInnerMessage;
                    m_pInnerMessage = 0;
                    return false;
                }
                break;
            case R_ELEMENT_LIST:
                // Decode Element List
                RElementList rInnerElementList;
                DecodeElementList(rInnerElementList, pszBuffer, uiBufferLength, strSrcAddress, strDestAddress, _vsContents, true);
                break;
            case R_FIELD_LIST:
                // Decode Field List
                DecodeFieldList(pszBuffer, uiBufferLength);
                break;
            case R_DATE:
                {
                    // Decode DATE
                    RDate rDate;
                    RRet = rDecodeDate( &m_rDecodeIter, &rDate );
                    if ( RRet != RET_SUCCESS )
                    {
                        return false;
                    }
                    std::stringstream sstream;
                    sstream << static_cast<int>( rDate.day ) << "/" << static_cast<int>( rDate.month ) << "/" << rDate.year;
                    _vsContents.push_back(sstream.str());
                }
                break;
            case R_DATETIME:
                {
                    // Decode DATETIME
                    RDateTime rDateTime;
                    RRet = rDecodeDateTime( &m_rDecodeIter, &rDateTime );
                    if ( RRet != RET_SUCCESS )
                    {
                        return false;
                    }
                    RBuffer rStringBuffer;
                    RRet = rDateTimeToString( &rStringBuffer, R_DATETIME,  &rDateTime );
                    if ( RRet != RET_SUCCESS )
                    {
                        return false;
                    }
                    std::stringstream sstream;
                    sstream << static_cast<int>( rDateTime.date.day ) << "/" << static_cast<int>( rDateTime.date.month ) << "/" << static_cast<int>( rDateTime.date.year) << " " << static_cast<int>( rDateTime.time.hour )
                        << ":" << static_cast<int>( rDateTime.time.minute ) << ":" << static_cast<int>( rDateTime.time.second ) << "." << static_cast<int>( rDateTime.time.millisecond ); 
                    _vsContents.push_back(sstream.str());
                }
                break;
            case R_DOUBLE:
                // Decode DOUBLE
                RDouble rDouble;
                RRet = rDecodeDouble( &m_rDecodeIter, &rDouble );
                _vsContents.push_back(boost::lexical_cast<std::string>(rDouble));
                //m_sStringStream << rDouble << ",";
                break;
            case R_UINT:
                // Decode UINT
                RUInt rUInt;
                RRet = rDecodeUInt( &m_rDecodeIter, &rUInt );
                _vsContents.push_back(boost::lexical_cast<std::string>(rUInt));
                //m_sStringStream << rUInt << ",";
                break;
            case R_ASCII_STRING:
                {
                // Decode STRING
                RBuffer rStringBuffer;
                RRet = rDecodeBuffer( &m_rDecodeIter, &rStringBuffer );
                std::string strData(rStringBuffer.data);
                _vsContents.push_back(strData);
                //m_sStringStream << rStringBuffer.data << ",";
                }
                break;
            case R_NO_DATA:
                RRet = RET_SUCCESS;
                break;
            default:
                RRet = RET_FAILURE;
                break;
            }
        }
        std::stringstream ssReport;
        std::copy(_vsContents.cbegin(),_vsContents.cend(),std::ostream_iterator<std::string>(ssReport,","));
        vsContents.push_back(ssReport.str());
    }
    else
    {
        return false;
    }
    return true;
}

不幸的是，它必须按照这个顺序，因为存储的字符串向量将包含逗号分隔的元素列表，稍后我将输出到csv列表。这段代码只需解码某些元素，并将生成的字符串推回到字符串向量中，然后将其放入字符串流中。

有人对如何提高性能有什么建议吗？？？谢谢