From owner-freebsd-stable@FreeBSD.ORG Sat Sep 9 02:48:26 2006 Return-Path: X-Original-To: freebsd-stable@freebsd.org Delivered-To: freebsd-stable@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 0D2C116A403 for ; Sat, 9 Sep 2006 02:48:26 +0000 (UTC) (envelope-from verbo.solo@sbcglobal.net) Received: from smtp103.plus.mail.re2.yahoo.com (smtp103.plus.mail.re2.yahoo.com [206.190.53.28]) by mx1.FreeBSD.org (Postfix) with SMTP id 75CAB43D53 for ; Sat, 9 Sep 2006 02:48:25 +0000 (GMT) (envelope-from verbo.solo@sbcglobal.net) Received: (qmail 3335 invoked from network); 9 Sep 2006 02:48:24 -0000 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=sbcglobal.net; h=Received:Subject:From:To:Content-Type:Date:Message-Id:Mime-Version:X-Mailer:Content-Transfer-Encoding; b=YffbnjfHn1NkBBisMRO3EEwcQU3UugEWI6vRGkQezHh4MEgsArzlTiFzY6oKKtpN1hss6TrJPU4Ar2+kmPOzZ4kzO5010iKOMRISZCoX2Ln01SJrSd0MAIsRRBSCpCimoMFggy1hu6f8oVbjpdI70re/oHSLkR+kp6xR4yl8/U8= ; Received: from unknown (HELO elbereth.gateway.2wire.net) (verbo.solo@sbcglobal.net@68.94.90.134 with plain) by smtp103.plus.mail.re2.yahoo.com with SMTP; 9 Sep 2006 02:48:24 -0000 From: "Charles P. Schaum" To: freebsd-stable@freebsd.org Content-Type: text/plain Date: Fri, 08 Sep 2006 21:48:22 -0500 Message-Id: <1157770102.2117.15.camel@elbereth.gateway.2wire.net> Mime-Version: 1.0 X-Mailer: Evolution 2.6.3 FreeBSD GNOME Team Port Content-Transfer-Encoding: 7bit Subject: Re: Broken loader in STABLE X-BeenThere: freebsd-stable@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Production branch of FreeBSD source code List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 09 Sep 2006 02:48:26 -0000 > CFLAGS= -O2 -pipe -funroll-loops -ffast-math I had used similar optimizations soon after going from 6.1-RELEASE to 6-STABLE and discovered it was a bad idea. Using -O -pipe has since stood me in good stead. The system is by no means sluggish, even on a cut-rate build-on-a-budget machine. I experienced that tuning for FS bottlenecks has a far greater performance impact. Splitting cache across different drives can have a positive effect similar to putting /usr/src and /usr/obj on different drives. After finding out where my needs were and partitioning appropriately, I have few complaints even with minimal optimization. That also holds for NetBSD on my dodgy old ACPI plus early AGP box that causes FreeBSD to panic unless I yank the s3 Trio for a couple of older cards and disable ACPI (1997 Microstar EISA/PCI/AGP K6 "transitional" hardware). Even when I just slap on an extra 2 gig drive for cache and /usr/obj, Net runs noticeably faster. Charles