Date: Wed, 28 Nov 2007 17:19:42 -0500 (EST) From: Daniel Eischen <deischen@freebsd.org> To: "M. Warner Losh" <imp@bsdimp.com> Cc: arch@freebsd.org Subject: Re: Code review request: small optimization to localtime.c Message-ID: <Pine.GSO.4.64.0711281718530.24547@sea.ntplx.net> In-Reply-To: <20071128.151021.709401576.imp@bsdimp.com> References: <20071128.151021.709401576.imp@bsdimp.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 28 Nov 2007, M. Warner Losh wrote:
> Please find enclosed some small optimizations. I know they make a
> couple lines too long, I'll correct that before I commit. They make
> the time functions do less redundant locking.
>
> However, in many places we do the following:
>
> pthread_mutex_lock();
> if (!is_set) {
> is_set = true;
> // do something
> }
> pthread_mutex_unlock();
>
> This is wasteful. We get locks ALL the time for every time operation,
> when in fact we need it more rarely. If we can tolerate losing a
> race, we can eliminate the locking in all but the startup case and
> those threads racing the startup:
>
> if (!is_set) {
> pthread_mutex_lock();
> if (!is_set) {
> is_set = true;
> // do something
> }
> pthread_mutex_unlock();
> }
>
> here, we know that is_set only ever changes from false to true. If it
> is already true, there's nothing to do. If it is false, we may need
> to do something, so we lock, check to see if we really need to do it,
> etc.
>
> Can anybody see a flaw in this logic?
Is this not a good place to use pthread_once() instead?
--
DE
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.GSO.4.64.0711281718530.24547>