Re: [Boost-users] boost::mutex synchronization problem

Subject: Re: [Boost-users] boost::mutex synchronization problem
From: Anthony Williams (anthony.ajw_at_[hidden])
Date: 2009-05-05 16:32:14

Ovanes Markarian <om_boost_at_[hidden]> writes:

>> On Tue, May 5, 2009 at 12:42 PM, Anthony Williams <anthony.ajw_at_[hidden]>
>> wrote:
>>
>>     Zachary Turner <divisortheory_at_[hidden]> writes:
>>
>>     > I've been trying to figure out if there's a simple way using condition
>>     > variables and mutexes to design a function with the following properties
>>     > (cliffs notes first, and then longer version in case it's not enough
>>     info):
>>     >
>>     > 1) there is a section of code in the function such that two threads
>>     cannot
>>     > enter that section at the same time.
>>     > 2) after leaving the section of code, the thread should wait for
>>     something to
>>     > happen (signalled by another thread)
>>     > 3) the waits in step 2 should be awoken in the order they occured.
>>
>>     There's nothing in boost to do this. I remember a thread on
>>     comp.programming.threads about a FIFO mutex (which is what you seem to
>>     want). Someone provided an implementation (which may well do what you
>>     want), but the consensus seemed to be "don't do that". If your program
>>     behaviour depends on threads being woken in a specific order then it may
>>     be better to rewrite the code so it doesn't.

> Is not barrier suitable for that?

No. Barrier forces N threads to wait until all N have reached the
barrier, then releases them all.

Anthony
-- 
Author of C++ Concurrency in Action | http://www.manning.com/williams
just::thread C++0x thread library   | http://www.stdthread.co.uk
Just Software Solutions Ltd         | http://www.justsoftwaresolutions.co.uk
15 Carrallack Mews, St Just, Cornwall, TR19 7UL, UK. Company No. 5478976