From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 12122 invoked by alias); 12 Apr 2002 12:54:32 -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 11918 invoked from network); 12 Apr 2002 12:54:25 -0000 Received: from unknown (HELO gate2.knmi.nl) (145.23.254.151) by sources.redhat.com with SMTP; 12 Apr 2002 12:54:25 -0000 Received: (from fwmaster@localhost) by gate2.knmi.nl (8.9.1a/8.6.12) id OAA00643; Fri, 12 Apr 2002 14:54:22 +0200 (MET DST) Received: by gate2p1.knmi.nl via smap (V1.3) id sma000434; Fri, 12 Apr 02 14:53:55 +0200 Received: from knmi.nl (bgwd73.knmi.nl [145.23.19.166]) by BCSXCS.knmi.nl with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2650.21) id 217YDPZQ; Fri, 12 Apr 2002 14:53:58 +0200 Message-ID: <3CB6D8E2.66047055@knmi.nl> Date: Fri, 12 Apr 2002 05:58:00 -0000 From: Toon Moene Organization: Royal Netherlands Meteorological Institute X-Accept-Language: en MIME-Version: 1.0 To: Andi Kleen CC: gcc@gcc.gnu.org, starlex@eng.sun.com Subject: Re: contribution References: <3CB6D448.45A4FE3E@knmi.nl.suse.lists.egcs> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-SW-Source: 2002-04/txt/msg00493.txt.bz2 Andi Kleen wrote: > > > You could compiled 172.mgrid with gcc twice this way and it showed > > > 48% improvement ? Or did you use a different compiler as backend? > > > If you used gcc - is it known where that huge difference comes > > > from? > To be honest I don't believe that he got that result with multiple runs > of gcc. If yes that would be a very bad bug in the optimizer. More likely > it was reached with the proprietary Sun compiler as backend (that would > make the patch more a porting tool for GNU C programs to Sun C) Hmm, I wouldn't be surprised at all ... When g77 came out (in February '95) we had a hard time beating the f2c+gcc combo. It might still be the case that, e.g., multi-rank array code (as in mgrid) is treated more optimally by the gcc middle-end when first "squashed" by a Fortran to C converter into a linear array. -- Toon Moene, KNMI, PO Box 201, 3730 AE De Bilt, The Netherlands. Tel. +31302206443, Fax +31302210407, e-mail moene@knmi.nl URL: http://www.knmi.nl/hirlam