实时进程间通讯问题 - 未名空间MITBBS历史存档

国际科技财经博客移民网络热点娱乐民生时事公众号

Redian新闻

>未名空间

>Programming - 葵花宝典

实时进程间通讯问题

实时进程间通讯问题# Programming - 葵花宝典

r*s2008-06-24 07:06

1 楼

【以下文字转载自 robotics 讨论区】
发信人: robotics (机器人技术), 信区: robotics
标题: 实时进程间通讯问题
发信站: BBS 未名空间站 (Tue Jun 24 23:45:44 2008), 转信
两个不同的real time进程运行在不同的CPU上面，使用一段共享内存进行同步
内存容量足够大，我想采用的方法，就是划分块
A块属于A进程写，B只读的区域，
B块属于B进程写，A只读的区域。
由于没有使用中断，两边都是在固定的时钟周期内写自己的区域同时读对方的区域
这样编程比较简单，我想了解一下可能出现的后果，
比如，A读了B正在写的区域
那么我可以为每个块设置一个标志位，代表在写，在读，或者ready
有没有哪位大侠给出一个简单但是确保安全的办法？

s*e2008-06-24 07:06

2 楼

如果不用任何lock的话，我觉得一个标志位可能不够，需要多个，比如要写，正在写，
写完。一点想法。

a*l2008-06-24 07:06

3 楼

I think that is basically a "lock" problem, and ususally it is not trivial
without any hardware support. On the other hand, you may consider some
operations will not be interrupted, e.g. , read/write to one bit. So,
starting from there, you can have something. E.g., one says "intend to write
", one follows "acknowledge to write", then one writes.

【在 r******s 的大作中提到】

: 【以下文字转载自 robotics 讨论区】
: 发信人: robotics (机器人技术), 信区: robotics
: 标题: 实时进程间通讯问题
: 发信站: BBS 未名空间站 (Tue Jun 24 23:45:44 2008), 转信
: 两个不同的real time进程运行在不同的CPU上面，使用一段共享内存进行同步
: 内存容量足够大，我想采用的方法，就是划分块
: A块属于A进程写，B只读的区域，
: B块属于B进程写，A只读的区域。
: 由于没有使用中断，两边都是在固定的时钟周期内写自己的区域同时读对方的区域
: 这样编程比较简单，我想了解一下可能出现的后果，

p*o2008-06-24 07:06

4 楼

Just FYI, for concurrency without ANY hardware support, Lamport's
bakery algorithm could be used ...
research.microsoft.com/users/lamport/pubs/pubs.html#bakery
Cited from the page:
"Before the bakery algorithm, people believed that the mutual exclusion prob
lem was unsolvable--that you could implement mutual exclusion only by using
lower-level mutual exclusion. Brinch Hansen said exactly this in a 1972 pap
er. Many people apparently still believe it. (See [90].) "

write

【在 a****l 的大作中提到】

: I think that is basically a "lock" problem, and ususally it is not trivial
: without any hardware support. On the other hand, you may consider some
: operations will not be interrupted, e.g. , read/write to one bit. So,
: starting from there, you can have something. E.g., one says "intend to write
: ", one follows "acknowledge to write", then one writes.

m*e2008-06-24 07:06

5 楼

Which CPU?

【在 r******s 的大作中提到】

r*s2008-06-24 07:06

6 楼

Good reference, Thanks a lot!

prob
using
pap
trivial

【在 p***o 的大作中提到】

: Just FYI, for concurrency without ANY hardware support, Lamport's
: bakery algorithm could be used ...
: research.microsoft.com/users/lamport/pubs/pubs.html#bakery
: Cited from the page:
: "Before the bakery algorithm, people believed that the mutual exclusion prob
: lem was unsolvable--that you could implement mutual exclusion only by using
: lower-level mutual exclusion. Brinch Hansen said exactly this in a 1972 pap
: er. Many people apparently still believe it. (See [90].) "
:
: write

j*p2008-06-24 07:06

7 楼

seqlock

【在 r******s 的大作中提到】

: Good reference, Thanks a lot!
:
: prob
: using
: pap
: trivial

l*g2008-06-24 07:06

8 楼

If you are using *NIX systems, POSIX IPC can give you sufficient supports:
Solution 1: message queue. but you will need thread support if two processes
running on different cycle times.
solution 2: semaphore + shares memory, but you need to make sure there is
no deadlock.
I am not sure how to do it on windows.

【在 r******s 的大作中提到】