From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 31592 invoked by alias); 9 Sep 2010 16:17:13 -0000 Received: (qmail 31584 invoked by uid 22791); 9 Sep 2010 16:17:12 -0000 X-SWARE-Spam-Status: No, hits=-1.9 required=5.0 tests=AWL,BAYES_00,RCVD_IN_DNSWL_NONE X-Spam-Check-By: sourceware.org Received: from vsmtp3.tin.it (HELO vsmtp3.tin.it) (212.216.176.223) by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Thu, 09 Sep 2010 16:17:04 +0000 Received: from [192.168.0.4] (79.53.13.46) by vsmtp3.tin.it (8.5.113) id 4BCE36060AB348D2; Thu, 9 Sep 2010 18:16:57 +0200 Message-ID: <4C890879.6060204@oracle.com> Date: Thu, 09 Sep 2010 16:39:00 -0000 From: Paolo Carlini User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.1.11) Gecko/20100714 SUSE/3.0.6 Thunderbird/3.0.6 MIME-Version: 1.0 To: Michael Matz CC: gcc-patches@gcc.gnu.org, Ira Rosen Subject: Re: Fix PR43432, vectorize backwards stepping loads References: <4C88F19F.7020708@oracle.com> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit 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: 2010-09/txt/msg00843.txt.bz2 Hi, Thanks Michael for the clarifications. > I'm not too familiar with the STL algorithms, but from a quick glance the > only *_backward things are move_backward, copy_backward and > merge_backward. Those just seem to merely copy items around, not actually > doing something with them. But they do have also the stores stepping > backwards, so the patch would need extending to support that. > Right, this is something I was missing: I'm pretty sure we have some pretty straightforward patterns in some cases, but definitely backward stores too. Unless you have, for example, a std::accumulate feed reverse_iterators?!? In fact *any* algorithm becomes backward with reverse_iterators, but those aren't that common, I'm afraid... Paolo.