From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-return-49687-listarch-gcc=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 9783 invoked by alias); 9 Apr 2002 13:17:27 -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 9701 invoked from network); 9 Apr 2002 13:17:24 -0000
Received: from unknown (HELO mail.cs.tu-berlin.de) (130.149.17.13)
  by sources.redhat.com with SMTP; 9 Apr 2002 13:17:24 -0000
Received: from platon.cs.tu-berlin.de (matzmich@platon.cs.tu-berlin.de [130.149.25.107])
	by mail.cs.tu-berlin.de (8.9.3/8.9.3) with ESMTP id PAA07512;
	Tue, 9 Apr 2002 15:14:46 +0200 (MET DST)
Received: (from matzmich@localhost)
	by platon.cs.tu-berlin.de (8.9.3/8.9.3) id PAA28202;
	Tue, 9 Apr 2002 15:14:45 +0200 (MET DST)
Date: Tue, 09 Apr 2002 06:21:00 -0000
From: Michael Matz <matzmich@cs.tu-berlin.de>
X-X-Sender:  <matzmich@platon>
To: "David S. Miller" <davem@redhat.com>
cc: <gcc@gcc.gnu.org>
Subject: Re: Undocumented anomaly with EXTRA_CONSTRAINTS
In-Reply-To: <20020409.050715.75514317.davem@redhat.com>
Message-ID: <Pine.GSO.4.33.0204091503080.18092-100000@platon>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
X-SW-Source: 2002-04/txt/msg00339.txt.bz2

Hi,

On Tue, 9 Apr 2002, David S. Miller wrote:

> These extra:
>
> 	1) accept any pseudo if !REG_OK_STRICT
> 	2) accept only non-renumbered pseudos if REG_OK_STRICT
>
> checks in their memory EXTRA_CONTRAINT entries aren't documented.
>
> Anyways, the gist of my email is:
>
> 1) Am I right?

I'm not quite sure what this has to do with only EXTRA_CONSTRAINT.  Any
'm' constraint accepts pseudos before reload and unassigned pseudos while
reloading.  Exactly because unassigned pseudos _are_ memory slots.

> 2) Once decided, we should document this and perhaps even
>    "help" the port somehow.
>
> By "help" the port, I mean we can somehow hide these recog/reload
> details about accepting pseudo-registers for memory constraints.

Hide from what?  From the .md files?  I.e. that patterns having 'm' in the
constraints not seeing REG rtx's?  This would mean to substitute
stackslots into those REG rtx before calling any recognizing things, which
is not what reload want to do.  In a later pass it might get a hardreg, so
replacing a pseudo with a mem too soon would be bad.


Ciao,
Michael.