linux-mips
[Top] [All Lists]

Re: [PATCH] Optimize generic get_unaligned / put_unaligned implementatio

To: Ralf Baechle <ralf@linux-mips.org>
Subject: Re: [PATCH] Optimize generic get_unaligned / put_unaligned implementations.
From: Jeremy Fitzhardinge <jeremy@goop.org>
Date: Thu, 15 Feb 2007 15:05:36 -0800
Cc: Andrew Morton <akpm@linux-foundation.org>, Atsushi Nemoto <anemo@mba.ocn.ne.jp>, linux-mips@linux-mips.org, linux-kernel@vger.kernel.org
In-reply-to: <20070215221839.GA14103@linux-mips.org>
Original-recipient: rfc822;linux-mips@linux-mips.org
References: <20050830104056.GA4710@linux-mips.org> <20060306.203218.69025300.nemoto@toshiba-tops.co.jp> <20060306170552.0aab29c5.akpm@osdl.org> <20070214214226.GA17899@linux-mips.org> <20070214203903.8d013170.akpm@linux-foundation.org> <20070215143441.GA18155@linux-mips.org> <20070215135358.020781dd.akpm@linux-foundation.org> <20070215221839.GA14103@linux-mips.org>
Sender: linux-mips-bounce@linux-mips.org
User-agent: Thunderbird 1.5.0.9 (X11/20070212)
Ralf Baechle wrote:
> Gcc info page says:
>
> [...]
> `packed'
>      The `packed' attribute specifies that a variable or structure field
>      should have the smallest possible alignment--one byte for a
>      variable, and one bit for a field, unless you specify a larger
>      value with the `aligned' attribute.
> [...]
>
> Qed?

So that the compiler has to assume that if its accessing this __packed
structure, it may be embedded unaligned within something else? And
because the pointer is cast through (void *) it isn't allowed to use
alias analysis to notice that the pointer wasn't originally (apparently)
unaligned.

Seems sound to me.

    J

<Prev in Thread] Current Thread [Next in Thread>