From mboxrd@z Thu Jan  1 00:00:00 1970
From: Albert ARIBAUD <albert.u.boot@aribaud.net>
Date: Wed, 17 Apr 2013 21:23:26 +0200
Subject: [U-Boot] [PATCH] crc32: Correct endianness of crc32 result
In-Reply-To: <CAPnjgZ1mR3h5fqCrhAuj4wN_4YCD4iO_VBnM76rHupnpowXHZg@mail.gmail.com>
References: <1365203470-9099-1-git-send-email-sjg@chromium.org>
	<20130406070429.381092002D9@gemini.denx.de>
	<CAPnjgZ2jRVpQ56_EpVKUMH8E1L2LJ0HgzNCs-GrN9bXfOSXz+Q@mail.gmail.com>
	<20130417054047.9D5692001AD@gemini.denx.de>
	<CAPnjgZ1mR3h5fqCrhAuj4wN_4YCD4iO_VBnM76rHupnpowXHZg@mail.gmail.com>
Message-ID: <20130417212326.4f803bfb@lilith>
List-Id: <u-boot.lists.denx.de>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: u-boot@lists.denx.de

Hi Simon -- and sorry for the dupe.

On Wed, 17 Apr 2013 11:28:07 -0700, Simon Glass <sjg@chromium.org>
wrote:

> I tried using:
> 
> #ifdef USE_HOSTCC
>    crc = htobe32(crc);
> #else
>    crc = cpu_to_be32(crc);
> #endif
>    memcpy(output, &crc, sizeof(crc));
> 
> 
> This is one instruction (4 bytes, 16%) smaller, but I suspect quite a
> lot slower due to the overhead of a very small memcpy().
> 
> 43e2c1d8: e28d1008 add r1, sp, #8
> 43e2c1dc: e3a02004 mov r2, #4
> 43e2c1e0: e6bf0f30 rev r0, r0
> 43e2c1e4: e5210004 str r0, [r1, #-4]!
> 43e2c1e8: e1a00004 mov r0, r4
> 43e2c1ec: eb001af7 bl 43e32dd0 <memcpy>

How about replacing the memcpy with an explicit put_unaligned(),
similar to what was done in

http://www.mail-archive.com/u-boot at lists.denx.de/msg109555.html

with get_unaligned()? The code will be longer than above, but shorter
than the above plus the memcpy(), and faster too -- actually, I'm
surprised that the compiler does not unroll the memcpy() on its own,
considering the size argument is a constant.

> Regards,
> Simon

Amicalement,
-- 
Albert.