From owner-svn-src-all@FreeBSD.ORG Sat Jul 2 19:05:34 2011 Return-Path: Delivered-To: svn-src-all@FreeBSD.org Received: by hub.freebsd.org (Postfix, from userid 1233) id 7A8031065673; Sat, 2 Jul 2011 19:05:34 +0000 (UTC) Date: Sat, 2 Jul 2011 19:05:34 +0000 From: Alexander Best To: Ulrich =?iso-8859-15?Q?Sp=F6rlein?= , Doug Barton , Warner Losh , Robert Watson , Joel Dahl , src-committers@FreeBSD.org, svn-src-all@FreeBSD.org, svn-src-head@FreeBSD.org Message-ID: <20110702190534.GA48471@freebsd.org> References: <201106110908.p5B98kkE066709@svn.freebsd.org> <4DF3B12C.8020505@FreeBSD.org> <9583277B-6EC0-4B7F-9640-2B128A96C6B8@bsdimp.com> <4DF3D633.5060506@FreeBSD.org> <20110702183320.GB8485@acme.spoerlein.net> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <20110702183320.GB8485@acme.spoerlein.net> Cc: Subject: Re: svn commit: r222980 - in head/sys: amd64/conf i386/conf X-BeenThere: svn-src-all@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: "SVN commit messages for the entire src tree \(except for " user" and " projects" \)" List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 02 Jul 2011 19:05:34 -0000 On Sat Jul 2 11, Ulrich Spörlein wrote: > On Sat, 11.06.2011 at 13:55:15 -0700, Doug Barton wrote: > > On 6/11/2011 1:02 PM, Warner Losh wrote: > > > > > > On Jun 11, 2011, at 12:17 PM, Doug Barton wrote: > > > > > >> On 6/11/2011 6:07 AM, Robert Watson wrote: > > >>> To me, this seems like the wrong direction. Over the last decade, we've > > >>> been trying to move away from conditional compilation of features to > > >>> having them be loadable as modules. > > >> > > >> FWIW, I agree. I'm wondering though, is there still a performance penalty for modules? My understanding in the past was that there is, although for most use cases it's in the statistical noise. Is that still true? > > > > > > At run time, I believe that's true. At load time, lots of modules can take a few seconds longer. > > > > I have 3 or 4 modules loaded via loader.conf at boot time. They take at > > least 2 seconds each. IMO loading everything via loader.conf would slow > > the boot so much as to be a non-starter. > > > > OTOH, I could imagine an rc.d script that depends on mountcritlocal that > > could load a list of modules. Unless I'm missing something that would be > > several times faster. > > I suspect this is your BIOS' fault. I load 22 modules via loader.conf > and the loader takes 2, at most 3, seconds to load them all (next to the > kernel). This is true for all machines that I own/owned. actually i recenly switched from using kernel modules to one big kernel entity, simply because i had the same problem. each module was taking ~ 3-5 seconds to load. make that 10 modules and you have additional loading time of almost half a minute. now that i'm using one big kernel entity i was able to reduce the loading time to ~ 4 seconds plus an extra 3 seconds for nvidia.ko. i could upload two videos demonstrating the huge differences between both schemes on my computer, if somebody is interested in it. cheers. alex > > As you can guess, I'm very much in favour of moving modules from GENERIC > to loader.conf ... > > Uli