From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 16925 invoked by alias); 18 May 2007 22:32:53 -0000 Received: (qmail 16913 invoked by uid 22791); 18 May 2007 22:32:51 -0000 X-Spam-Check-By: sourceware.org Received: from smtp-out.google.com (HELO smtp-out.google.com) (216.239.45.13) by sourceware.org (qpsmtpd/0.31) with ESMTP; Fri, 18 May 2007 22:32:48 +0000 Received: from zps78.corp.google.com (zps78.corp.google.com [172.25.146.78]) by smtp-out.google.com with ESMTP id l4IMWd0W005557; Fri, 18 May 2007 15:32:40 -0700 Received: from localhost.localdomain.google.com (dhcp-172-18-118-205.corp.google.com [172.18.118.205]) (authenticated bits=0) by zps78.corp.google.com with ESMTP id l4IMWYks024588 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NOT); Fri, 18 May 2007 15:32:35 -0700 To: Eric Botcazou Cc: gcc-patches@gcc.gnu.org, Rask Ingemann Lambertsen , Uros Bizjak , Paolo Bonzini , Chris Lattner Subject: Re: [PATCH, x86_64]: Provide longlong.h definitions for 128bit operations References: <464B6834.1050600@gmail.com> <200705180912.20137.ebotcazou@libertysurf.fr> <200705182238.36051.ebotcazou@libertysurf.fr> From: Ian Lance Taylor Date: Fri, 18 May 2007 22:32:00 -0000 In-Reply-To: <200705182238.36051.ebotcazou@libertysurf.fr> Message-ID: User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.4 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-IsSubscribed: yes Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Archive: List-Post: List-Help: Sender: gcc-patches-owner@gcc.gnu.org X-SW-Source: 2007-05/txt/msg01251.txt.bz2 Eric Botcazou writes: > > For a non-CC0 style machine things are simpler. Just split multi-word > > operations before reload. There was never any particular need to wait > > until after reload to split them in the past. And now you really > > don't want to wait. > > Of course SPARC-V8 is the perfect counter-example since you have a pairing > instruction "ldd" for consecutive integer registers. :-) Yes, good point, you probably don't want to split a DImode load or store there. You can still split everything else, though, and the right thing should happen. Unfortunately this means that if one of the halves of the load is unused, we won't convert back to an SImode load. Ian