Date: Fri, 18 Nov 2016 11:58:01 +0100 From: Polytropon <freebsd@edvax.de> To: Matt Smith <matt.xtaz@gmail.com> Cc: "Christoph P.U. Kukulies" <kuku@kukulies.org>, "freebsd-questions@freebsd.org" <freebsd-questions@freebsd.org> Subject: Re: how long does a make buildworld normally take? Message-ID: <20161118115801.4ef94f0d.freebsd@edvax.de> In-Reply-To: <20161118101752.GA39102@gmail.com> References: <b3bc55ef-5389-0f46-752a-290a45ee97fc@kukulies.org> <20161118101752.GA39102@gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, 18 Nov 2016 10:17:52 +0000, Matt Smith wrote: > On Nov 18 10:38, Christoph P.U. Kukulies wrote: > >how long does a make buildworld normally take? > >Has make buildworld time constant over the years? > > > > > >In 1995 "make world" on a Pentium Pro about 18 hours > >On my 4GB, CPU: AMD Athlon(tm) 64 X2 Dual Core Processor 3800+ > >(1995.64-MHz 686-class CPU) > >make buildworld took about 18 hours. Well that CPU isn't the latest > >either (10 Years or so) > > > > 18 hours?! On my current home-server which is an Intel Atom D525 1.8ghz > it only takes 5 hours. And these are not known for being fast at all. Allow me to quote from a message I wrote in 2008, in order to illustrate technical progress by collectingg several build times for different world and kernel configurations. FreeBSD 5 on Pentium 4 with 2 GHz and 1 GB RAM: b.world+b.kern: 17494.415u 2562.134s 5:46:42.25 96.4% (with CFLAGS) 17474.169u 2481.368s 5:46:30.40 95.9% (without CLFAGS) 5608.712u 1595.130s 2:13:18.67 90.0% 6382.185u 1788.433s 2:26:36.06 92.8% buildworld: 5086.993u 1431.086s 1:58:16.33 91.8% 11457.047u 2151.158s 3:54:15.31 96.8% buildkernel 2326.380u 234.457s 43:42.15 97.6% 1102.491u 278.194s 25:18.58 90.9% 1182.203u 294.622s 26:12.71 93.9% 1518.402u 310.741s 34:16.96 88.9% 3289.368u 529.669s 1:05:25.90 97.2% installkernel: 5.718u 6.898s 0:30.97 40.6% 6.655u 7.389s 0:32.08 43.7% 6.994u 7.734s 0:33.19 44.3% (...software advance happens here...) FreeBSD 7 on Pentium 4 with 2 GHz and 1 GB RAM: b.world+b.kern: 16574.070u 2516.128s 6:06:03.90 86.9% (with debug) 18232.967u 2427.404s 7:19:49.24 78.2% (with debug) 18992.839u 2569.146s 9:12:00.28 65.1% buildworld: 11457.047u 2151.158s 3:54:15.31 96.8% buildkernel: 3289.368u 529.669s 1:05:25.90 97.2% 3503.732u 524.399s 1:11:05.53 94.4% 4032.019u 572.636s 1:58:29.08 64.7% (with debug) installkernel: 17.396u 12.587s 0:46.89 63.9% 18.890u 12.131s 1:11.85 43.1% As you can see, 5 hours was a possible value on a single-core single-threat slow-as-ass CPU. But then the system became more advanced, and 7 - 9 hours compile time became possible. :-) > In > 1996 ish I was doing it on something like a pentium 75 and I think it > still took about 5 hours. So you're right there! 18 hours seems way too > long though. On my first FreeBSD machine, compiling world and kernel needed approx. 24 hours. The kernel itself could be rebuilt in more or less 6 hours. Today we have loadable modules. :-) > These days it compiles things like clang which take a very long time. But it also introduces optimization which the "old compilers" did not provide, so by investing more build time, you get a better runtime experience (programs run faster). Especially on limited hardware, this is worth considering (like in ye olde days setting several "strange" CFLAGS to make mplayer perform better). -- Polytropon Magdeburg, Germany Happy FreeBSD user since 4.0 Andra moi ennepe, Mousa, ...
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20161118115801.4ef94f0d.freebsd>