Date: Mon, 2 Jun 2008 17:59:46 +0100
From: Alan Cox <alan@lxorguk.ukuu.org.uk>
To: lsorense@csclub.uwaterloo.ca (Lennart Sorensen)
Cc: Rodolfo Giometti <giometti@enneenne.com>,
       Andrew Morton <akpm@linux-foundation.org>, linux-kernel@vger.kernel.org
Subject: Re: LinuxPPS low-level IRQs timestamps & ldisc
Message-ID: <20080602175946.366c5d64@core>
In-Reply-To: <20080602170916.GD16164@csclub.uwaterloo.ca>
References: <20080601161510.GA26854@enneenne.com>
	<20080602075113.60d0fdf8@core>
	<20080602092551.GC26854@enneenne.com>
	<20080602150257.54692f35@core>
	<20080602145603.GG26854@enneenne.com>
	<20080602170916.GD16164@csclub.uwaterloo.ca>
Organization: Red Hat UK Cyf., Amberley Place, 107-111 Peascod Street,
 Windsor, Berkshire, SL4 1TE, Y Deyrnas Gyfunol. Cofrestrwyd yng Nghymru a
 Lloegr o'r rhif cofrestru 3798903
Mime-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8BIT
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 877
Lines: 18

> Of course if you have to check of the DCD changed, that may not be
> constant.  A check for whether this device currently has anyone that
> wants the timestamp though should be a constant time.

Not having to take the cost of that for other users is far more important
than a time variance of a jump prediction. CPU cache effects alone will
add way more variance already. It's not a meaningful number to chase
compared to the randomness introduced by the IRQ handling itself.

If you want perfect accuracy do what I²IT did years ago and use a card
which timestamps the physical IRQ bus transition and exposes it to the OS.

Alan
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/