当前位置:网站首页>2020-11-06: go, let's talk about the scheduler.

2020-11-06: go, let's talk about the scheduler.

2020-11-06 22:43:54 Fuda Dajia architect's daily question

Fogo's answer 2020-11-06:
·MPG Model :goroutine The concurrency model of can be summarized as MPG Model ;
·MPG Concept : Threads (machine, System threads , Physical threads )- kernel (processor)- coroutines (goroutine, User threads , Logical threads );
· Many to many scheduling model : The overall scheduling follows the many to many model , Multiple collaborations ( About a million ) Scheduling in multiple threads at the same time ( About a thousand ) Next ;
·LRQ(LocalRunningQueue): Local run queue , One M Execute in a P In the context of ,P There is a G queue ;
·GRQ(GlobalRunningQueue): Global run queue ,G Not initialized or not LRQ When it's ready to mount, it's thrown in GRQ;
·GRQ=>LRQ:MP Will be in LRQ Check after execution GRQ, And steal the task from it and mount it to the current LRQ In the implementation of , They also have regular checks ;
· The purpose of scheduling is : The purpose of scheduling is to prevent thread blocking 、 idle 、 By OS Hang up (syscall);
· Prevent threads M jam :G1 coroutines IO When you leave MP,G2 From the current MP Of LRQ Pop up and execute ;
· Prevent threads M idle :M1 When idle , From M2 Of LRQ To steal half of the mission , Mount to your own LRQ In the implementation of ;
· Prevent threads M By OS Hang up (syscall):P With LRQ Hang to the bottom of other threads to execute , When syscall At the end ,M Will try to G0 Mount to other LRQ Medium or GRQ in ;


detailed go Scheduler model reference :
Go Chinese language network :https://studygolang.com/articles/9610
Simple books :https://www.jianshu.com/p/f9024e250ac6

版权声明
本文为[Fuda Dajia architect's daily question]所创,转载请带上原文链接,感谢