[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <48FD19E9.8010807@flurg.com>
Date: Mon, 20 Oct 2008 18:53:13 -0500
From: "David M. Lloyd" <dmlloyd@...rg.com>
To: david@...g.hm
Cc: Arnaldo Carvalho de Melo <acme@...hat.com>,
linux-kernel <linux-kernel@...r.kernel.org>
Subject: Re: sched_yield() options
On 10/20/2008 06:08 PM, david@...g.hm wrote:
> in the case I'm looking at there are two (or more) threads running with
> one message queue in the center.
>
> 'input threads' are grabbing the lock to add messages to the queue
>
> 'output threads' are grabbing the lock to remove messages from the queue
>
> the programmer is doing a pthread_yield() after each message is
> processed in an attempt to help fairness (he initially added it in when
> he started seeing starvation on single-core systems)
>
> what should he be doing instead?
If you're seeing starvation, to me that's a good indicator that the
granularity of queue items are too small... probably there'd be an overall
benefit of grabbing more things at once from the queue.
To make a silly metaphor out of it - imagine you've got a bunch of office
interns (threads) whose jobs are to fill out PQ9-12b forms. An intern has
to wait in line before they can get to the desk where the secretary is
handing out the forms to be filled out, one by one. An intern can fill out
the form just as fast as the secretary hands it to them; if they do so,
then one intern ends up doing all the work while the others just wait in line.
But in order to make everyone equally busy (by injecting sched_yield())
you're making the interns take the paper, walk back to their desk, fill it
out, and then get back in line, which takes somewhat more time overall,
despite making the interns look much busier. A better solution would be to
give each intern a big stack of forms so that they spend a greater
proportion of time filling them out rather than waiting in line; perhaps if
each intern is kept busy enough, the line will always be empty, which would
be the best situation of all.
- DML
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists