From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-return-61654-listarch-gcc=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 13094 invoked by alias); 15 Oct 2002 21:37:22 -0000
Mailing-List: contact gcc-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Archive: <http://gcc.gnu.org/ml/gcc/>
List-Post: <mailto:gcc@gcc.gnu.org>
List-Help: <http://gcc.gnu.org/ml/>
Sender: gcc-owner@gcc.gnu.org
Received: (qmail 13087 invoked from network); 15 Oct 2002 21:37:21 -0000
Received: from unknown (HELO bjl1.asuk.net) (81.29.64.88)
  by sources.redhat.com with SMTP; 15 Oct 2002 21:37:21 -0000
Received: from bjl1.asuk.net (localhost [127.0.0.1])
	by bjl1.asuk.net (8.12.5/8.12.5) with ESMTP id g9FLdtVn032445;
	Tue, 15 Oct 2002 22:39:55 +0100
Received: (from jamie@localhost)
	by bjl1.asuk.net (8.12.5/8.12.5/Submit) id g9FLdtt5032443;
	Tue, 15 Oct 2002 22:39:55 +0100
X-Authentication-Warning: bjl1.asuk.net: jamie set sender to egcs@tantalophile.demon.co.uk using -f
Date: Tue, 15 Oct 2002 15:32:00 -0000
From: Jamie Lokier <egcs@tantalophile.demon.co.uk>
To: Kevin Lawton <kevinlawton2001@yahoo.com>
Cc: Michael Matz <matz@suse.de>, gcc@gcc.gnu.org
Subject: Re: Request of new __attribute__ for switch statements (elimination of the bounds check)
Message-ID: <20021015213954.GC31811@bjl1.asuk.net>
References: <20021015200247.GA31811@bjl1.asuk.net> <20021015201332.21851.qmail@web80311.mail.yahoo.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20021015201332.21851.qmail@web80311.mail.yahoo.com>
User-Agent: Mutt/1.4i
X-SW-Source: 2002-10/txt/msg00878.txt.bz2

Kevin Lawton wrote:
> The __attibute_ angle is simple, I can exactly tell the
> compiler when I care to have it applied (maybe only one
> case in all the code), and it disappears quickly with
> an #ifdef'd macro.  Which makes it play nice with "other"
> compilers.

If your proposal is actually implemented, could it please be changed
so that the attribute goes _inside_ the switch parentheses?

Then it can be used to optimise "variant type" macros without having
to uglify thousands of switch statements in a program (just the macro
definitions).  To use a few example from GCC itself, these kind of
can all be optimised in this way:

    switch (TREE_CODE (t)) { /* ... */ }
    switch (GET_CODE (rtx)) { /* ... */ }
    switch (GET_MODE (rtx)) { /* ... */ }

(I'm not sure you'd want to optimise these in GCC because of the risk
of bugs, but there are programs which use similar macros where it is
more desirable to optimise).

-- Jamie