Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 17 Aug 2002 04:25:44 +1000 (EST)
From:      Bruce Evans <bde@zeta.org.au>
To:        Poul-Henning Kamp <phk@critter.freebsd.dk>
Cc:        David Greenman-Lawrence <dg@dglawrence.com>, Alfred Perlstein <bright@mu.org>, <cvs-committers@FreeBSD.org>, <cvs-all@FreeBSD.org>
Subject:   Re: cvs commit: src/sys/kern uipc_socket2.c 
Message-ID:  <20020817041512.K8159-100000@gamplex.bde.org>
In-Reply-To: <14228.1029507987@critter.freebsd.dk>

next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, 16 Aug 2002, Poul-Henning Kamp wrote:

> In message <20020816235317.I7073-100000@gamplex.bde.org>, Bruce Evans writes:
>
> >Not unless a very raw timestamp method were used.  Using nanotime()
> >would add a 10(?)% overhead to some syscalls even if the hardware part
> >took no time.  Something using rdtsc() in syscall() might be fast enough,
> >but this would give similar problems for scaling of very large counts
>
> The scaling issue could possibly be dealt with using a periodic (1
> Hz) function which does the scaling and accumulation in timeval
> format.

I guess doing that for all processes would be not much worse than
schedcpu() (schedcpu() would be a good place to do it).  Is schedcpu()
ever too inefficient with any actual number of processes on current
machines?

Bruce


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020817041512.K8159-100000>