Why is epoll efficient?

Epoll The bottom implementation of is shown in the figure above ：（ understand ）

• （ Red and black trees , Is a special binary tree ,Epoll How can I monitor many ？ First create the root node of the tree , Each root node is a fd Store in the form of a structure （ The node contains some attributes , contain callback function ）, For a tree, you can attach nodes at will .
• Linked list , When a file descriptor generates an event , Automatically called callback function , Through the callback callback Function to find the event corresponding to the linked list （ Read time or write events ）, The linked list is an event linked list .

epoll Why efficient ？

1.select and poll The monitored files need to be copied and transferred between the operating system and the kernel layer , Each poll needs to be copied from the user layer to the kernel layer , Efficiency is very low , and epoll The kernel layer and the user layer share memory , The monitor event descriptor does not exist in the kernel layer nor in the application layer , The kernel layer monitors directly from the shared space , Greatly improved efficiency

2.select Polling mechanism used , Polling , That is, the event occurrence query will be performed for each monitored event descriptor , Low efficiency , and epoll Use the event notification mechanism , namely

The event will be notified , Instead of polling , Greatly improved efficiency .