Date: Thu, 11 Apr 2002 12:09:02 -0400 (EDT) From: Kenneth Culver <culverk@alpha.yumyumyum.org> To: Michael Lucas <mwlucas@blackhelicopters.org> Cc: Storms of Perfection <ancient@outloud.org>, <freebsd-hackers@FreeBSD.ORG> Subject: Re: FreeBSD Advanced tuning advice Message-ID: <20020411115814.Y33533-100000@alpha.yumyumyum.org> In-Reply-To: <20020411110538.A93751@blackhelicopters.org>
next in thread | previous in thread | raw e-mail | index | archive | help
> Yes, but every time I've seen those brought up, the high-level hackers > here say "Don't do that." :) > > If we could get a definitive answer on acceptable flags, I'll put it > in the FAQ. > Basically (although I'm not a "high-level hacker") what I've gathered is that -O optimization above -O (without a numbeR) isn't supported because there are known bugs in gcc opts above -O (and it's questionable whether or not they actually improve performance). Also, I know in a lot of cases the -march opts don't really do much either. In some cases doing -march=pentium actually produces faster code on current systems than doing -march=pentiumpro (or -march=i686, these are the same thing). So basically my recommendation is that no matter what you hear from linux people (just saw an article on slashdot, and the guy said he compiled the whole gentoo linux system with -march=i686, which really doesn't increase performance much ) say about these opts, they don't really help. Ken To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020411115814.Y33533-100000>