From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 29318 invoked by alias); 24 Oct 2004 13:02:03 -0000 Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Archive: List-Post: List-Help: Sender: gcc-patches-owner@gcc.gnu.org Received: (qmail 29310 invoked from network); 24 Oct 2004 13:02:02 -0000 Received: from unknown (HELO smtp2.fuse.net) (216.68.8.175) by sourceware.org with SMTP; 24 Oct 2004 13:02:02 -0000 Received: from gx4.fuse.net ([66.42.186.147]) by smtp2.fuse.net (InterMail vM.6.01.03.04 201-2131-111-106-20040729) with ESMTP id <20041024130151.BAPD21290.smtp2.fuse.net@gx4.fuse.net> for ; Sun, 24 Oct 2004 09:01:51 -0400 Received: from dellpi.pinski.fam ([66.42.186.147]) by gx4.fuse.net (InterMail vG.1.00.00.00 201-2136-104-20040331) with ESMTP id <20041024130153.BULP3628.gx4.fuse.net@dellpi.pinski.fam>; Sun, 24 Oct 2004 09:01:53 -0400 Received: from [10.0.0.80] (zhivago.i.pinski.fam [10.0.0.80]) by dellpi.pinski.fam (8.12.2/8.12.1) with ESMTP id i9OD1tVT000863; Sun, 24 Oct 2004 09:01:56 -0400 (EDT) In-Reply-To: <20041024094112.GA1374@atrey.karlin.mff.cuni.cz> References: <20041024094112.GA1374@atrey.karlin.mff.cuni.cz> Mime-Version: 1.0 (Apple Message framework v618) Content-Type: text/plain; charset=US-ASCII; format=flowed Message-Id: Content-Transfer-Encoding: 7bit Cc: gcc-patches@gcc.gnu.org From: Andrew Pinski Subject: Re: [patch] Partial fix for PR 18048 Date: Sun, 24 Oct 2004 17:30:00 -0000 To: Zdenek Dvorak X-SW-Source: 2004-10/txt/msg02012.txt.bz2 On Oct 24, 2004, at 5:41 AM, Zdenek Dvorak wrote: > Hello, > > this patch contains several improvements to ivopts and related > optimizations that improve the code produced for mgrid benchmark quite > a > bit (although even with the patch, ivopts still spoil the code). > > The improvements are: > > 1) Ivopts produce expressions like > > *(&a[start] + 8 * index) > > and believe that the memory access created for it will include the > multiplication by 8. This fails if there are several similar > accesses, since dom will CSE the "8 * index" part (thus adding > unnecessary multiplication instruction, and increasing register > pressure). Isn't this already done by fold_stmt? -- Pinski