From owner-freebsd-amd64@FreeBSD.ORG Mon Jul 4 05:13:20 2005 Return-Path: X-Original-To: amd64@freebsd.org Delivered-To: freebsd-amd64@FreeBSD.ORG Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 5601916A41C; Mon, 4 Jul 2005 05:13:20 +0000 (GMT) (envelope-from chat95@mac.com) Received: from smtpout.mac.com (smtpout.mac.com [17.250.248.87]) by mx1.FreeBSD.org (Postfix) with ESMTP id 4092E43D46; Mon, 4 Jul 2005 05:13:20 +0000 (GMT) (envelope-from chat95@mac.com) Received: from mac.com (smtpin01-en2 [10.13.10.146]) by smtpout.mac.com (Xserve/8.12.11/smtpout05/MantshX 4.0) with ESMTP id j645DJan017614; Sun, 3 Jul 2005 22:13:19 -0700 (PDT) Received: from localhost ([133.11.172.102]) (authenticated bits=0) by mac.com (Xserve/smtpin01/MantshX 4.0) with ESMTP id j645DFs2006054; Sun, 3 Jul 2005 22:13:19 -0700 (PDT) Date: Mon, 04 Jul 2005 14:12:57 +0900 (JST) Message-Id: <20050704.141257.104030539.chat95@mac.com> To: amd64@freebsd.org From: NAKATA Maho Organization: private X-Mailer: Mew version 3.3 on Emacs 21.3 / Mule 5.0 (SAKAKI) Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: netchild@freebsd.org Subject: better optimization flags for opteron (amd64) X-BeenThere: freebsd-amd64@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Porting FreeBSD to the AMD64 platform List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 04 Jul 2005 05:13:20 -0000 Better optimization flags for Opteron (amd64) Recently AMD released a guide http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/32035.pdf Compiler Usage Guidelines for 64-Bit Operating Systems on AMD64 Platforms it has nice flag settings. recommended flags for opteron/gcc-3.4 is followings: -pipe -O3 -ffast-math -funroll-all-loops -fpeel-loops -ftracer -funswitch-loops -funit-at-a-time (see p.19 of 32035.pdf) and my benchmark of math/sdpara on Opteron 1.6GHz dual shows: * scalapack/mpich/blacs is not optimized 2004.59 real 1979.83 user 9.77 sys * scalapack/mpich/blacs is optimized 1261.85 real 1242.63 user 7.76 sys Hope this helps for you. -- NAKATA, Maho (maho@FreeBSD.org)