From owner-freebsd-current Wed Dec 1 5: 7:44 1999 Delivered-To: freebsd-current@freebsd.org Received: from not.demophon.com (vpn.iscape.fi [195.170.146.67]) by hub.freebsd.org (Postfix) with ESMTP id E639814D48 for ; Wed, 1 Dec 1999 05:07:40 -0800 (PST) (envelope-from will@not.demophon.com) Received: (from will@localhost) by not.demophon.com (8.9.3/8.8.7) id PAA54461; Wed, 1 Dec 1999 15:05:25 +0200 (EET) (envelope-from will) To: Bruce Evans Cc: current@freebsd.org, tejblum@arc.hq.cti.ru Subject: Re: kernel: -mpreferred-stack-boundary=2 ?? References: From: Ville-Pertti Keinonen Date: 01 Dec 1999 15:05:25 +0200 In-Reply-To: Bruce Evans's message of "1 Dec 1999 09:57:48 +0200" Message-ID: <86d7squ7ze.fsf@not.demophon.com> Lines: 20 X-Mailer: Gnus v5.5/XEmacs 20.4 - "Emerald" Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Bruce Evans writes: > If the caller has passed a double, then the stack alignment can be determined > at compile time and the current subl-type adjustment can be used. Doubles are not aligned when passed as parameters and passing a double doesn't guarantee that there are local variables in the caller that are doubles. Anyhow, I don't think significant changes to gcc should be made in FreeBSD when they only affect code size and performance, suggestions to improve such things should be discussed on the gcc list and adopted through official gcc releases. The defaults can, of course, be changed, but userland code is not that size-critical. In any case, userland code should at least use double-alignment, otherwise you could get poor performance in common floating point code. SSE code is not necessarily worth that much consideration yet. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message