无锁机制的循环缓冲技术

 

此技术是借鉴了Linux内核源代码技术，无锁应用条件是同时只有一个写线程和读线程访问缓冲区时，不用加互斥锁，因为读写两个计数器各不影响，除此以外，还是要加互斥锁，如写线程只有一个线程，读线程有N个同时访问，那么写端不用加，只在读端加互斥锁即可。

主要就两个方法：
unsigned int CRingBuffer::GetBuffer(unsigned char *pBuffer, unsigned int iSize)
{
unsigned int iLen;
iSize = min(iSize, m_WritePtr - m_ReadPtr);
/* first get the data from m_ReadPtr until the end of the pBuffer */
iLen = min(iSize, m_Size - (m_ReadPtr & (m_Size - 1)));
memcpy(pBuffer, m_Buffer + (m_ReadPtr & (m_Size - 1)), iLen);
/* then get the rest (if any) from the beginning of the pBuffer */
memcpy(pBuffer + iLen, m_Buffer, iSize - iLen);
m_ReadPtr += iSize;
return iSize;
}
unsigned int CRingBuffer:PutBuffer(unsigned char *pBuffer, unsigned int iSize)
{
unsigned int iLen;
iSize = min(iSize, m_Size - m_WritePtr + m_ReadPtr);
/* first put the data starting from m_WritePtr to pBuffer end */
iLen = min(iSize, m_Size - (m_WritePtr & (m_Size - 1)));
memcpy(m_Buffer + (m_WritePtr & (m_Size - 1)), pBuffer, iLen);
/* then put the rest (if any) at the beginning of the pBuffer */
memcpy(m_Buffer, pBuffer + iLen, iSize - iLen);
m_WritePtr += iSize;
return iSize;
}

只要保证一个线程读，一个线程写，就不用锁。
其实质就是让某一端保持一个线程在操作，例如：

1、如果写端只有一个线程，读线程有N个同时访问，那么写端不用加，只在读端加互斥锁即可;
2、如果读端只有一个线程，写线程有N个同时访问，那么读端不用加，只在写端加互斥锁即可。