Date: Mon, 2 Apr 2007 14:18:56 -0700 From: "Steven Hillis" <evultrole@gmail.com> To: "Royce Williams" <royce@alaska.net>, freebsd-sparc64@freebsd.org Subject: Re: freebsd-update/sparc64 and buildworld statistics Message-ID: <340594530704021418tdbf2358sce77a8ecb054d584@mail.gmail.com> In-Reply-To: <46100531.90604@alaska.net> References: <340594530702211416l1a9d03k8051fc5d685f841b@mail.gmail.com> <46100531.90604@alaska.net>
next in thread | previous in thread | raw e-mail | index | archive | help
> > Sorry for the long pause on this thread - Thunderbird's thread display > was collapsed and I didn't see the additional posts. > > >> I'd still like to see what a quad Ultra 80 can do, though. :) Who has > > access to one of these? > > > > I have one of these under my foot right now. 4x450 4GB memory, 2xScsi > > drives. I went ahead a did a fresh install to get rid of all the weird > > Thanks very much! Were /usr and /usr/obj on different filesystems > and/or different disks? Yes, /usr and /usr/obj are on different disks, noatime and async set on usr/obj > Standard "/usr/bin/time -h make -j5 buildworld" was somewhere around 1h45m > > (sorry, I didn't write it down the first time), so that's probably the > best > > you'll get. > > 1h45m is terrific compared to some of the others, but I can't say that > I have a stastically significant sample yet, by any means. > > > > However, after enabling ccache, while the first run was roughly the same > > (couple minutes longer), the second run (after caching) gave these > results: > > > > 49m14.74s real 1h28m39.70s user 1h4m8.45s sys > > Added to my worldstone statistics: > > http://www.alaska.net/~royce/freebsd/sparc64/worldstone/ > <http://www.alaska.net/%7Eroyce/freebsd/sparc64/worldstone/>; > > > > So, that sort of breaks the 1 hour limit? I can't imagine there will > enough > > source changes between updates to add 10 minutes to the compile time, > but I > > hear people have had some trouble with ccache and buildworld so whether > it > > solves the problem or not seems sort of up in the air. > > I've written Colin to ask his opinion. Can anyone else comment on the > ccache/world interaction problem? This thread seems to indicate that > some issues have been resolved: > > http://lists.freebsd.org/pipermail/freebsd-current/2006-March/061839.html I've certainly had far less problems with ccache than trying to use the noclean options. To date I haven't had any issues using it (on 4 different architectures), I think it is just when you try to customize things after the first build that people have headaches... which makes no sense given that it works off of the preprocessor, so any changes should show up as different... but that's the only time I've had a problem with it... > dmesg, if you care: > > real memory = 4294967296 (4096 MB) > > What branch were you tracking? I try to note that in both the dmesg > and worldstone info. In the meantime, I've added your dmesg to the > archive as u80_unknown_unknown. At the time it was simply the most up to date 6.2 branch, guess I missed that part in the copy/paste... I have nothing better to do later, so I'll see how the time on current goes as well
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?340594530704021418tdbf2358sce77a8ecb054d584>