From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 24214 invoked by alias); 17 Jan 2004 15:14:20 -0000 Mailing-List: contact gcc-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Archive: List-Post: List-Help: Sender: gcc-owner@gcc.gnu.org Received: (qmail 24170 invoked from network); 17 Jan 2004 15:14:19 -0000 Received: from unknown (HELO dberlin.org) (69.3.5.6) by sources.redhat.com with SMTP; 17 Jan 2004 15:14:19 -0000 Received: from [192.168.1.7] (account dberlin [192.168.1.7] verified) by dberlin.org (CommuniGate Pro SMTP 4.1.4) with ESMTP-TLS id 5870957; Sat, 17 Jan 2004 10:14:18 -0500 In-Reply-To: <20040117145355.GA15982@deneb.enyo.de> References: <10401170324.AA15949@vlsi1.ultra.nyu.edu> <1074309907.3147.139.camel@frodo.toronto.redhat.com> <20040117134753.GA14621@deneb.enyo.de> <42BD9D55-48FC-11D8-AECF-000A95DA505C@dberlin.org> <20040117145355.GA15982@deneb.enyo.de> Mime-Version: 1.0 (Apple Message framework v609) Content-Type: text/plain; charset=US-ASCII; format=flowed Message-Id: Content-Transfer-Encoding: quoted-printable Cc: "gcc@gcc.gnu.org" From: Daniel Berlin Subject: Re: [RFC] Contributing tree-ssa to mainline Date: Sat, 17 Jan 2004 15:14:00 -0000 To: Florian Weimer X-SW-Source: 2004-01/txt/msg01064.txt.bz2 On Jan 17, 2004, at 9:53 AM, Florian Weimer wrote: > Daniel Berlin wrote: > >> As Diego stated when SRA was committed, the following limitation=20 >> exists: >> /* Maximum number of fields that a structure should have to be >> scalarized. >> FIXME This limit has been arbitrarily set to 5. Experiment to=20 >> find >> a >> sensible setting. */ >> #define MAX_NFIELDS_FOR_SRA 5=03 >> >> This was done, IIRC, to avoid severely increased register pressure. >> >> If you want SRA to do something on your structure, you'll need to >> increase that limit > > If I go down to two fields/tuples of length two, the optimization > doesn't kick in, either. 8-( Hmmm. Interesting. Can you file a bug report with this testcase (the length 2, which=20 *should* be taken care of already), so it can be tracked? --Dan