- 基于c++内存池,共享内存和信号量实现高速的进程间通信队列,支持单进程读单进程写,单进程多进程写,多进程读单进程写,多进程读多进程写
- 把消息队列对象生成在实现分配好的共享内存区中(共享内存去的大小远远大于消息队列管理对象messagequeue),对象中记录者共享内存去
剩余内存的数据情况,消息内存区为一个环形内存区。如下图:
原理图
详细介绍, - 写的时候移动end索引,读的时候移动begin索引,保证了在单进程读和单进程写的时候是线 程安全的,多进程读多进成写时利用信号量集(一个读信号和写信号)实现进程间的共享内存读写锁来保证多进程安全。
- 源码main.cpp
- 测试结果:
===============================================================
Try to malloc share memory of 10240 bytes...
Successfully alloced share memory block, (key=12345), id = 11272200, size = 10240
Mem trunk address 0x0x7ffff7ff4000,key 11272200 , size 10240, begin 0, end 0, queue module 0
Write SingleRWTest thread 1 ,write count 1000000
Read SingleRWTest ,thread 1 ,read count 1000000
Touch to share memory key = 12345...
Now remove the exist share memory 11272200
Remove shared memory(id = 11272200, key = 12345) succeed.
=======================SingleRWTest=============================
SingleRWTest cost time 3046 ms
Read read_count 1000000
Write write_count 1000000
SingleRWTest ok 0
===============================================================
Try to malloc share memory of 10240 bytes...
Successfully alloced share memory block, (key=1234), id = 11304977, size = 10240
Mem trunk address 0x0x7ffff7ff1000,key 11304977 , size 10240, begin 0, end 0, queue module 3
===============================================================
Write MulRWTest thread 4 ,write count 100000
Write MulRWTest thread 1 ,write count 100000
Write MulRWTest thread 0 ,write count 100000
Write MulRWTest thread 3 ,write count 100000
Write MulRWTest thread 2 ,write count 100000
Read MulRWTest ,thread 3 ,read count 99981
Read MulRWTest ,thread 1 ,read count 104765
Read MulRWTest ,thread 2 ,read count 104513
Read MulRWTest ,thread 4 ,read count 95279
Read MulRWTest ,thread 0 ,read count 95467
Touch to share memory key = 1234...
Now remove the exist share memory 11304977
Remove shared memory(id = 11304977, key = 1234) succeed.
=======================MulRWTest===============================
MulRWTest cost time 24639 ms
Read read_count 500000
Write write_count 500000
MulRWTest ok - 为了操作方便,这里的读写操作用全部用线程代替进程,从目的上来说效果是一样的
- 优先考虑单进程读单进程写,无锁读写效率更快,多进程读在一定程度上会降低收发效率,注意这里所说的多进程读写,如果有两个进程 一个进程只读不写一个进程只写不读,此时为单进程读单进程写,而不是多进程读写
- 如果数据来自网络,注意字节序列,大小端的问题
- 进程意外退出时可以不卸载共享内存区,重新启动时共享内存中的消息不会丢失,会重新因射到内存中, 可以继续读取
- 创建共享内存队列时,注意读写模式
- ONE_READ_ONE_WRITE, //一个进程读消息一个进程写消息(推荐)
- ONE_READ_MUL_WRITE, //一个进程读消息多个进程写消息
- MUL_READ_ONE_WRITE, //多个进程读消息一个进程写消息
- MUL_READ_MUL_WRITE, //多个进程读消息多个进程写消息