Date: Sun, 16 Nov 1997 21:27:13 -0800 From: "Jordan K. Hubbard" <jkh@time.cdrom.com> To: Amancio Hasty <hasty@rah.star-gate.com> Cc: "John S. Dyson" <toor@dyson.iquest.net>, sprice@hiwaay.net (Steve Price), smp@csn.net, current@FreeBSD.ORG Subject: Re: make world time???/ Message-ID: <28686.879744433@jkh.cdrom.com> In-Reply-To: Your message of "Sun, 16 Nov 1997 17:21:31 PST." <199711170121.RAA00618@rah.star-gate.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> I do a make world in about 75 minutes -- I just didn't do profiled libraries. Argh. If we're going to compare worldstone numbers, we've really got to converge on some standard for what consititutes a proper one. ;-) In my own benchmarks, I always do a completely "standard" build with only one exception, which is to add CFLAGS=-O -pipe to /etc/make.conf (oh yeah, /usr also mounted async). Other folks stripe /usr/obj or mount it noatime or do any number of other things (like NOPROFILE) which can only skew the numbers when comparing them. Maybe the first step would be to agree upon a script for launching the world build which sets various environment variables consistently? Then we'd at least be down to just the I/O trickieries as variants. I've definitely noticed, with my dual P6/233 (and yes Virginia, it is possible to overclock a P6/180 to a P6/233 if you have good fans and some luck :), that things are primarily *I/O* bound, not CPU bound, with my single IBM DCAS 4.3 GB drive (these are 5400 RPM and not quite speed demons). On a machine with /usr/src and /usr/obj mounted on a 5 drive (Quantum 2GB) CCD array, I can shave as much as 40 minutes off the world build just on a uniprocessor P6/200, which is definitely food for thought. Jordan
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?28686.879744433>