From mboxrd@z Thu Jan  1 00:00:00 1970
From: Wolfgang Denk <wd@denx.de>
Date: Thu, 02 Dec 2010 19:51:36 +0100
Subject: [U-Boot] [PATCH] omap3evm: Clean-up EVM detection code.
In-Reply-To: <4CF7C7F4.6030803@free.fr>
References: <1291288812-12653-1-git-send-email-premi@ti.com>
	<20101202113726.2C683EA6600@gemini.denx.de>
	<4CF7896B.5090007@free.fr>
	<20101202120141.30416EA6600@gemini.denx.de>
	<4CF7922B.3020504@free.fr>
	<20101202135804.3BACFEA6600@gemini.denx.de>
	<4CF7C7F4.6030803@free.fr>
Message-ID: <20101202185136.499ACEA6600@gemini.denx.de>
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

Dear Albert ARIBAUD,

In message <4CF7C7F4.6030803@free.fr> you wrote:
>
> > Well, an u8 is as good a data type as any other. The available range
> > of 0...255 seems more than sufficient to store the needed
> > information, so why should I waste 4 bytes of storage when a single
> > byte is sufficient as well?
> 
> You don't necessarily use only one byte when declaring an u8 instead of 
> an int, because the next declaration may have alignment requirements 
> that will cause the compiler to skip bytes after the u8. Besides, u8 is 

The compiler / linker may (or may not) optimize this and collect
variables of similar alignment. An "int foo;" is likely to end in
.bss segment, while an "char foo;" will probably show up in .sbss - I
don;t know how good or bad the current situation for ARM is, but I'm
sure it is improving (look for example at all the microoptimizations
done by Linaro).

> not "as good a data type" as any other, it is a specific data type 
> whereas 'int' is the native data type of the platform, supposed to be 
> the most natural to deal with for the cpu -- 32-bit for an ARM.

Can an ARM CPU not read1s and write single bytes, too?

> u8 are perfect and normal, for instance, as fields of a structure which 
> represents byte registers, or to perform 8-bit arithmetic. Here, 
> however, there is indeed no reason to use any specific type, so we 
> should use the cpu's native type.

I do not share your opinion.

But this is a pretty academic topic, and I'm neither in the mood nor
do I have the time for lengthy discussions.  Let's stop this here.

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Cigarette, n.: A fire at one end, a fool at the other, and a  bit  of
tobacco in between.