X-Authentication-Warning: delorie.com: mailnull set sender to djgpp-bounces using -f
From: Ben Pfaff <blp AT cs DOT stanford DOT edu>
Newsgroups: comp.os.msdos.djgpp,comp.lang.c
Subject: Re: gcc 2.953 output - is this a bug?
Date: 06 Feb 2002 18:39:29 -0800
Lines: 29
Sender: blp AT pfaff DOT stanford DOT edu
Message-ID: <87d6zim2vy.fsf@pfaff.stanford.edu>
References: <3C61E6BB DOT 7F9AD526 AT yahoo DOT com>
NNTP-Posting-Host: pfaff.stanford.edu
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.1
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

CBFalconer <cbfalconer AT yahoo DOT com> writes:

> Please look at the following disassembly:
> 
>       h = h * 37UL + (unsigned char) *string++;
>  585:   8d 04 d2                lea    (%edx,%edx,8),%eax <<***
>  588:   8d 04 82                lea    (%edx,%eax,4),%eax <<***
> 
> look at the lines marked "<<***".  I fail to see how this
> implements h * 37UL.

Why, you mean it's not perfectly obvious? :-)
Here's a translation back to C of these lines:
        eax = edx + 8 * edx
        eax = edx + 4 * eax
Here's an equivalent for the first line:
        eax = 9 * edx
which, performing a substitution on eax, means that the second
line becomes equivalent to
        eax = edx + 4 * (9 * edx)
and from there it should be evident.

(GCC is more clever than I thought.)

[snippage]
-- 
"It would be a much better example of undefined behavior
 if the behavior were undefined."
--Michael Rubenstein