Date: Wed, 13 Jan 2010 12:56:17 +0100
From: Andi Kleen <andi@firstfloor.org>
To: William Allen Simpson <william.allen.simpson@gmail.com>
Cc: Linux Kernel Developers <linux-kernel@vger.kernel.org>,
       Linux Kernel Network Developers <netdev@vger.kernel.org>,
       Ilpo =?iso-8859-1?Q?J=E4rvinen?= <ilpo.jarvinen@helsinki.fi>,
       Eric Dumazet <eric.dumazet@gmail.com>, Andi Kleen <andi@firstfloor.org>
Subject: Re: [PATCH v4] tcp: harmonize tcp_vx_rcv header length assumptions
Message-ID: <20100113115617.GA24818@basil.fritz.box>
References: <4B49D001.4000302@gmail.com> <4B4DA4F4.6060007@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <4B4DA4F4.6060007@gmail.com>
User-Agent: Mutt/1.5.17 (2007-11-01)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 997
Lines: 24

On Wed, Jan 13, 2010 at 05:48:20AM -0500, William Allen Simpson wrote:
> Harmonize tcp_v4_rcv() and tcp_v6_rcv() -- better document tcp doff
> and header length assumptions.
>
> Reduces multiply/shifts, marginally improving speed.
>
> Removes redundant tcp header length checks before checksumming.

I wonder if this actually improves performance on x86. On x86 and several
other architecture there's a addressing mode which allows to scale 
numbers by small factors (like 4). So doing a *4 is very cheap.

That's likely cheaper than using another register for the scaled value,
especially on 32bit x86 which doesn't have many.

It's difficult to benchmark this code, but did you check that the code 
shrinks after applying this patch at least?

-andi
--
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/