Subj : Re: Futexes are wrong! (Are they?)
To   : comp.programming.threads
From : Jomu
Date : Mon Jun 06 2005 01:08 pm



Marcin 'Qrczak' Kowalczyk wrote:
> "Jomu" <dragisha@gmail.com> writes:
>
> > As consumer-producer message queues are (for me, probably elsewhere
> > also) most frequent sync devices, and their problem with preemptive
> > signaling happens mostly (maybe only) when one thread signals while
> > holding mutex (same one waiting thread will try to reacquire when
> > signaled) then why don't rearrange signaling so that it happens
> > after mutex is freed?
>
> I guess it will work in this case.
>
> Unfortunately in the case I have pthread_cond_{broadcast,signal}
> is followed by pthread_cond_wait, so I can't move signalling after
> unlocking and additional context switches are pointless.

  You have code sample of this?

>
> BTW, is it better to use pthread_cond_broadcast or pthread_cond_signal
> when I know that only one thread is waiting?

  As preempting happens in any case, I see no reason for difference.
Only reason for broadcast would be if more than one thread fullfils
can-work after event whose completion you signal.

>
> --
>    __("<         Marcin Kowalczyk
>    \__/       qrczak@knm.org.pl
>     ^^     http://qrnik.knm.org.pl/~qrczak/

.