Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 27 Sep 2005 19:17:49 +1000
From:      Peter Jeremy <PeterJeremy@optushome.com.au>
To:        Kris Kennaway <kris@obsecurity.org>
Cc:        freebsd-current@freebsd.org
Subject:   Re: 4BSD/ULE numbers...
Message-ID:  <20050927091749.GT40237@cirb503493.alcatel.com.au>
In-Reply-To: <20050926224453.GB39901@xor.obsecurity.org>
References:  <200509261847.35558@harrymail> <20050926174738.GA57284@xor.obsecurity.org> <43387811.1090308@freebsd.org> <20050926224453.GB39901@xor.obsecurity.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 2005-Sep-26 18:44:53 -0400, Kris Kennaway wrote:
>I'd love a way to measure sched_lock contention..I'm sure it's a
>factor on e.g. my machines with >=10 CPUs, but mutex profiling doesn't
>see it because it's a spinlock.

How about adding a spincount field in struct mtx and incrementing it in
the inner loop of _mtx_lock_spin()?  This wouldn't be perfect but you
could get a rough idea of the lock contention by looking at the rate
that the spincount field incremented.

Working out how many CPUs are simultaneously contending for sched_lock
would be more difficult but doable.
-- 
Peter Jeremy



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050927091749.GT40237>